* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download CTL.SC4x – Technology and Systems
Survey
Document related concepts
Transcript
CTL.SC4x – Technology and Systems Key Concepts Document This document contains the Key Concepts for the SC4x course. These are meant to complement, not replace, the lesson videos and slides. They are intended to be references for you to use going forward and are based on the assumption that you have learned the concepts and completed the practice problems. This draft was created by Dr. Alexis Bateman in the spring of 2017. This is a draft of the material, so please post any suggestions, corrections, or recommendations to the Discussion Forum under the topic thread “Key Concept Documents Improvements.” Thanks, Chris Caplice, Eva Ponce and the SC4x Teaching Community Spring 2017 v1 V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 1 Table of Contents IntroductiontoDataManagement.......................................................................................................4 DataManagement..................................................................................................................................4 QueryingtheData...................................................................................................................................5 DataModeling.....................................................................................................................................7 RelationalModels....................................................................................................................................7 DesigningDataModels...........................................................................................................................8 RelationshipsandCardinality..................................................................................................................8 Keys.........................................................................................................................................................9 DatabaseNormalization.....................................................................................................................11 ObjectivesofNormalization..................................................................................................................11 Summaryoffivenormalforms..............................................................................................................12 ClientServerArchitecture......................................................................................................................13 DatabaseQueries...............................................................................................................................14 StructuredQueryLanguage...................................................................................................................14 CreatingDatabasesandTables.............................................................................................................16 DatabaseConditional,Grouping,andJoins........................................................................................18 DatabaseConditionalClauses...............................................................................................................18 SortingandSamplingData....................................................................................................................19 JoiningMultipleTables..........................................................................................................................21 IntroductiontoMachineLearning......................................................................................................23 OverviewofMachineLearningAlgorithms...........................................................................................23 ModelQuality........................................................................................................................................24 MachineLearningAlgorithms............................................................................................................26 Dimensionalityreduction.......................................................................................................................26 Principalcomponentanalysis(PCA)......................................................................................................26 Clustering..............................................................................................................................................27 Classifications........................................................................................................................................27 Comparingpredictoraccuracy..............................................................................................................29 Sensitivityandspecificity.......................................................................................................................29 SupplyChainSystems-ERP................................................................................................................31 SupplyChainITSystems........................................................................................................................31 EnterpriseResourcePlanning................................................................................................................31 ERPCommunication..............................................................................................................................33 TheValueofERPforSCM......................................................................................................................34 SupplyChainSystems-SupplyChainModules...................................................................................36 AdvancedPlanningSystems..................................................................................................................37 TransportationManagementSystems(TMS)........................................................................................38 ManufacturingExecutionSystems........................................................................................................41 V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 2 SupplyChainSystems-SoftwareSelection&Implementation...........................................................42 Architecture...........................................................................................................................................42 CloudComputing...................................................................................................................................42 SoftwareVendorSelection....................................................................................................................45 Implementation.....................................................................................................................................46 TechnologyTrends.............................................................................................................................48 Trend:AutonomousVehicles.................................................................................................................48 Trend:DeliveryDrones..........................................................................................................................48 Trend:MobileComputing......................................................................................................................49 AdditiveManufacturing........................................................................................................................49 V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 3 Introduction to Data Management Summary Supplychainsaremovingatever-fasterrateswithtechnologyandsystemssupportingthis movement.Datamanagementplaysacriticalroleinenablingsupplychainstomoveatthe speedandprecisiontheydotoday,andtheneedforadvanceddatamanagementwillonly continue. Inrecentyearstherehasbeenanexplosionofinformationandthisisespeciallytrueinsupply chains.AfewexamplesintroducedincludeAmazon’smassivesupplychainsselling480million uniqueitemsto244millioncustomerswhileUPSisdelivering20millionpackagesto8.4million deliverypoints.Thisinformationiscomingfrommultiplesources,inadditiontosensors,the “internetofthings”,andregulationsrequiringincreasingamountsofinformation. Allofthisinformationiscommonlyreferredtoasthe“BigData”challenge.Dataisdrivingour modernworld,buthowcanwebesureofitanduseitmosteffectively?Aswewillreview– dataismessy,itrequirescleaningandprogramming.Dataisfrequentlytrappedinsiloes comingfromdifferentsources,whichmakesworkingwithitmorechallenging.Inaddition,data isbigandgettingevenbiggerdaily.Thetoolswehaveallbecomecomfortablewith (spreadsheets)cannolongerhandlethatamountofdata,sowemustusedifferenttoolsto enablegreateranalysis. Tobetterunderstandtheroleofdataandhowtomanageit,thefollowingsummariescoveran introductiontodatamanagement,datamodeling,anddatanormalization–togetusstartedon asolidgroundwithhandlinglargedatasets–anabsoluteessentialinsupplychain management. Data Management Indatamanagementsupplychainmanagerswillbefacedwithimmensecomplexity.This complexityisinfluencedbythevolume(howmuch),velocity(pace),variety(spread),and veracity(accuracy).Eachofthesecomponentswillinfluencehowdataistreatedandusedin thesupplychain. Thereareseveralreoccurringissuesthatsupplychainmanagersmustbeawareofastheyare workingwithdata: • Isthedataclean? • Isthedatacomplete? • Whatassumptionsareyoumakingaboutthedata? • Aretheresultsmakingsense?HowcanIcheck? V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 4 Cleaningdataisoneofthemostimportant;yettimeconsumingprocessesindataanalysis.It cangreatlyinfluencetheoutcomeofanalysisifnotcompletedproperly.Therefore–SC professionalsshouldalwaysplanenoughtimeforbasicdatachecks(meaningifyougetgarbage in,youwillgetgarbageout). Thereareseveraltypicalchecksyoushouldalwayslookfor: • Invalidvalues-negative,text,toosmall,toobig,missing • Mismatchesbetweenrelateddatasets-#ofrows,#ofcols • Duplication–uniqueidentifiers • Humanerror–wrongdates,invalidassumptions • Alwaysexploretheoutliers–theyarethemostinteresting! Whencleaningdata,youshouldbeorganized.Thismeansyoumustmakesuretoversionthe documentsyouareworkingwithandkeeptrackofdatachanges. Querying the Data Onceyouhaveacleanandorganizedsetofdata,queryingthedatacanmakedataextremely powerful.Queryingdatareferstotheactionofretrievingdatafromyourdatabase.Becausea databasecanbesolarge–weonlywanttoqueryfordatathatfitsacertaincriteria. Thereareseveralbasicoptionsthatcanhelpyougetsomequickanswersinbigdatasets,such asusingPivotTables: • TherearedatasummarizationtoolsfoundinLibreOffice,GoogleSheets,andExcel • Theyautomaticallysort,count,totaloraveragethedatastoredinonetableor spreadsheet,displayingtheresultsinasecondtableshowingthesummarizeddata. • Veryusefulintabulatingandcross-tabulatingdata No more spreadsheets! Unfortunately,aswedivedeeperintothebigdatachallenge,wefindthatspreadsheetscanno longerserviceallofourneeds.Wehavethechoiceofworkingwithstructuredorunstructured data.Adatabaseisastructuredwayofstoringdata.Youcanimposerules,constraintsand relationshipsonit.Furthermoreitallowsfor: • Abstraction:Separatesdatausefromhowandwherethedataisstored.Thisallows systemstogrowandmakesthemeasiertodevelopandmaintainthroughmodularity. • Performance:Databasemaybetunedforhighperformanceforthetaskthatneedsto bedone(manyreads,manywrites,concurrency) Spreadsheetsareunstructureddata.Youhaveadatadumpintoonspreadsheetandyouneed tobeabletodolotsofdifferentthings.Spreadsheetswillalwaysbegreatforalimitedsetof analysissuchasinformal,causal,andone-offanalysisandprototyping.Unfortunatelytheyare nolongersuitedforrepeatable,auditable,orhighperformanceproduction.Unstructureddata commonlyhasproblemswith:redundancy,clarity,consistency,security,andscalability. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 5 Learning Objectives • • • • Understandtheimportanceofdatainsupplychainmanagement. Reviewtheimportanceofhighqualityandcleandatabases. Recognizethepowerofqueryingdata. Differentiatebetweenunstructuredandstructureddataandtheneedfortoolsbeyond spreadsheets. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 6 Data Modeling Summary Nowthatwehavebeenintroducedtodatamanagementandtheissueofbigdata,wenow deepdriveintodatamodelingwherewelearnhowtoworkwithdatabases.Datamodelingis thefirststepindatabasedesignandprogrammingtocreateamodelforhowdatarelatesto eachotherwithinadatabase.Datamodelingistheprocessoftransitioningalogicalmodelinto aphysicalschema. Tounderstandtheprocessofdatamodeling,wereviewseveralcomponentsincluding relationaldatabases,dataorganization,datamodelsfordesigningdatabases,andwhat constitutesagooddatamodel.Adatamodelconsistsofseveralpartsincluding:entitiesand attributes,primarykeys,foreignkeys,andrelationshipsandcardinality. Relational Models Therelationalmodelisanapproachtomanagingdatathatusesstructureandlanguagewhere alldataisgroupedintorelations.Arelationalmodelprovidesamethodforspecifyingdataand queries.Itisbasedonfirst-orderpredicatelogic,whichwasdescribedbyEdgarF.Coddin1969. Thislogicdefinesthatalldataisrepresentedintermsoftuples,groupedintorelations.There areseveraldefinitionstobefamiliarwithaswereviewedpreviouslywithrelationalmodels: • Entity:object,conceptorevent • Attribute(column):acharacteristicofanentity • Recordortuple(row):thespecificcharacteristicsorattributevaluesforoneexampleof anentity • Entry:thevalueofanattributeforaspecificrecord • Table:acollectionofrecords • Database:acollectionoftables Tables and Attributes Datainrelationaltablesareorganizedintotables,whichrepresententities.Singletableswithin adatabasecanbeseenassimilartoaspreadsheet.However,weusedifferentwordstoreferto “rows”and“columns”.Attributesarethecharacteristicsofanentity. Tables • Tablesrepresententities,whichareusuallypluralnouns • Tablesareoftennamedaswhattheyrepresent(typicallypluralnouns,withoutspaces): e.g.Companies,Customers,Vehicles,Orders,etc. Attributes • Characteristicsofanentity(table),typicallynouns • Examplesintheformof:Table(Attr1,Attr2,...AttrN),Vehicles(VIN,Color,Make, Model,Mileage) V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 7 EntityTypesandEntityoccurrence:anentityisanyobjectinthesystemwewanttomodeland store.Anentityoccurrenceisauniquelyidentifiableobjectbelongingtoanentitytype. Designing Data Models Thereareseveralstepstodesigningadatabasetostoreandanalyzedata. 1. Developadatamodelthatdescribesthedatainthedatabaseandhowtoaccessit 2. Datamodeldefinestablesandattributesinthedatabase(eachimportantconcept/noun inthedataisdefinedasatableinthedatabase) Datamodelshelpspecifyeachentityinatableinastandardizedway.Theyallowtheuserto imposerules,constraintsandrelationshipsonthedatathatisstored.Italsoallowsusersto understandbusinessrulesandprocessandanalyzedata. Rules for a Relational Data Model Thereareseveralrulesforrelationaldatamodel: • Actsasaschematicforbuildingthedatabase • Eachattribute(column)hasauniquenamewithinatable • Allentriesorvaluesintheattributeareexamplesofthatattribute • Eachrecord(row)isuniqueinagooddatabase • Orderingofrecordsandattributesisunimportant Whatmakesagoodrelationaldatamodel?Agoodrelationalmodelshouldbecompletewithall thenecessarydatarepresented.Thereshouldbenoredundancy.Businessrulesshouldbe effectivelyenforced.Modelsshouldalsobereusablefordifferentapplications.Andfinally,it shouldbeflexibleandbeabletocopewithchangestobusinessrulesordatarequirements. Relationships and Cardinality Whenwebegintoworkwiththedata–wehavetounderstandhowdatarelatestoeachother anddatauniquenessoftheattributes.Someofthiscanbemanagedthroughentitytypesand attributes.Relationships+cardinality=businessrules. Entity and Attributes Anentityisaperson,place,thing,orconceptthatcanbedescribedbydifferentdata.Each entityismadeofanumberofattributes.Entitytypesshouldbedescribedaspartofthedata modelingprocess,thiswillhelpwiththedocumentationanddeterminationofbusinessrules. Howtodrawandentity-relationshipdiagram: AnERDisagraphicalrepresentationofaninformationsystemthatvisualizestherelationship betweentheentitieswithinthatsystem. • ERDorentity-relationshipdiagramisaschematicofthedatabase • Entitiesaredrawnasboxes V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 8 • • Relationshipsbetweenentitiesareindicatedbylinesbetweentheseentities Cardinalitydescribestheexpectednumberofrelatedoccurrencesbetweenthetwo entitiesinarelationshipandisshownusingcrow'sfootnotation(seefiguresbelow) Cardinality–crow’sfootnotation GeneralMeanings Mandatoryvs.Optional DomainValidationEntities:Alsoknownaspicklistsorvalidationlists.Domainvalidation entitiesareusedtostandardizedatainadatabase,theyrestrictentriestoasetofspecified values.Theyaretableswithasingleattributethatenforcesvaluesofattributeinrelated table(s). Keys Primarykeysareattributesusedtouniquelyidentifyarecordwhileforeignkeysareattributes storedinadependententity,whichshowhowrecordsinthedependententityarerelatedto anindependententity. Primarykey:oneormoreattributesthatuniquelyidentifyarecord.Theattributehasbe uniquelysuited. ForeignKey:Primarykeyoftheindependentorparententitytypeismaintainedasanon-key attributeintherelated,dependentorchildentitytype,thisisknownastheforeignkey Compositekey:isaprimarykeythatconsistsofmorethanoneattribute,ex:charterairline, everyflighthasadifferentnumber. ManytoManyRelationships:Amanytomanyrelationshipreferstoarelationshipbetween tablesinadatabasewhenaparententitycontainsseveralchildentitytypesinthesecond table.ex-Vehiclecanbedrivenbymanydrivers,driverscandrivemanyvehicles.Inthiscasean associativetable(entity),akajunctiontableisappropriatewheretheprimarykeyofparentis usedinprimarykeyofchild. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 9 Referential integrity Referentialintegritymaintainsthevalidityofforeignkeyswhentheprimarykeyintheparent tablechanges.Everyforeignkeyeithermatchesaprimarykey(orisnull). Cascaderules:chooseamongdeleteoptions • Cascaderestrict:Rowsintheprimarykeytablecan’tbedeletedunlessallcorresponding rowsintheforeignkeytableshavebeendeleted • Cascadedelete:Whenrowsintheprimarykeytablearedeleted,associatedrowsin foreignkeytablesarealsodeleted Learning Objectives • • • • • • Thedatamodeldescribesthedatathatisstoredinthedatabaseandhowtoaccessit. Datamodelsenableuserstounderstandbusinessrulesandeffectivelyprocessand analyzedata. Understandthatbusinessrulesareimposedonthedatabasethroughrelationshipsand cardinality. Recognizethatdatamodelsmayvaryforagivendatasetasbusinesslogicevolves. Rememberthatthedatamodelingprocessmayrevealinconsistenciesorerrorsinthe data,whichwillhavetobecorrectedbeforeimportingintoadatabase. Selectionofentitiesandassociatedattributesfromaflatfileisnotalwaysobvious. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 10 Database Normalization Summary Databasenormalization,ornormalization,isanimportantstepindatabasemanagement. Normalizationisintrinsictorelationaldatabasesandistheprocessoforganizingattributesinto relations(ortables).Thisprocessisvitalinreducingdataredundancyandimprovingdata integrity.Inaddition,normalizationhelpsorganizeinformationaroundspecifictopicsthatcan beusedtodigestthemassiveamountofinformationindatabasesintosomethingdigestible. WhenSCprofessionalsarepresentedwithlargeamountsofrawdata,thatrawdatamaybe storedinasingletable,containingredundantinformationorinformationaboutseveral differentconcepts.Thedatacanbeseparatedintotablesandnormalizedtoallowforbetter datahandlingandcomprehension.Togettothisplace,updatingadatamodelcanbedone collaborativelyduringmeetingsanddiscussionstodefinethebusinessrules.Duringupdates, normalizationpreventsmistakesanddatainconsistencies.Normalizationhelpsprevent redundancy,confusion,improperkeys,wastedstorage,andincorrectoroutdateddata. Objectives of Normalization 1. Tofreethecollectionof[tables]fromundesirableinsertion,updateanddeletion dependencies. 2. Toreducetheneedforrestructuringthecollectionof[tables],asnewtypesofdataare introduced,andthusincreasethelifespanofapplicationprograms. 3. Tomaketherelationalmodelmoreinformativetousers. 4. Tomakethecollectionof[tables]neutraltothequerystatistics,wherethesestatistics areliabletochangeastimegoesby. **Rememberourrelationalmodeldefinitions • Entity:object,conceptorevent • Attribute(column):acharacteristicofanentity • Recordortuple(row):thespecificcharacteristicsorattributevaluesforoneexampleof anentity • Entry:thevalueofanattributeforaspecificrecord • Table:acollectionofrecords • Database:acollectionoftables V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 11 Summary of five normal forms 1. Allrowsinatablemustcontainthesamenumberofattributes;nosub-lists,no repeatedattributes. 2. Allnon-keyfieldsmustbeafunctionofthekey. 3. Allnon-keyfieldsmustnotbeafunctionofothernon-keyfields. 4. Arowmustnotcontaintwoormoreindependentmulti-valuedfactsaboutanentity. 5. Arecordcannotbereconstructedfromseveralsmallerrecordtypes. Normal Forms FirstNormalForm–thebasicobjectiveofthefirstnormalform(definedbyCodd)istopermit datatobequeriedandmanipulated,groundedinfirstorderlogic.Allrowsinatablemust containthesamenumberofattributes;nosub-lists,norepeatedattributes,identifyeachsetof relateddatawithaprimarykey.Firstnormalformcanmakedatabasesrobusttochangeand easiertouseinlargeorganizations. SecondNormalForms–mustfirstbeinfirstnormalform,allnon-keyfieldsmustbeafunction oftheprimarykey;onlystorefactsdirectlyrelatedtotheprimarykeyineachrow. ThirdNormalForm-mustfirstbeinsecondnormalform.Alltheattributesinatableare determinedonlybythecandidatekeysofthetableandnotbyanynon-primeattributes.Third normalformwasdesignedtoimprovedatabaseprocessingwhileminimizingstoragecosts. FourthNormalForm-mustfirstbeinthirdnormalform.Arowshouldnotcontaintwoormore independent,multi-valuedfactsaboutanentity.Fourthnormalformbeginstoaddressseveral issueswhenthereisuncertaintyinhowtomaintaintherows.Whentherearetwounrelated factsaboutanentity,theseshouldbestoredinseparatetables. FifthNormalForm-mustfirstbeinfourthnormalform.Arecordcannotbereconstructed fromseveralsmallerrecordtypes.Sizeofthissingletableincreasesmultiplicatively,whilethe normalizedtablesincreaseadditively.Mucheasiertowritethebusinessrulesfromthethree tablesinthefifthnormalform,rulesaremoreexplicit.Supplychainstendtohavefifthnormal formissues. Normalization Implementation Details Normalizationensuresthateachfactisstoredinoneandonlyoneplace,toensuredata remainsconsistent.Normalizingthedatamodelisatechnicalexercise.Itdoesnotchange businessrules!However,throughtheprocessofmeetingsanddecisionsitmayhelptherules befurtherdefinedthroughreview.Careindatanormalizationisneededtopreservedata quality.Therearetimeswhennormalizationisnotanoption–thishappenswhenthereare large,readonlydatabasesforreportgenerationofdatawarehouses. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 12 Client-servermodel Client Server Architecture Client Server Model Clientscanconnecttoserverstoaccessaspecificserviceusingastandardizedprotocol,see • Clientscanconnecttoserverstoaccessaspecificservice figurebelow. usingastandardizedprotocol Web Applica9on Client DatabaseUser InterfaceClient MySQL Database Server Analy9cs Client 46 Database Servers Databasesarehostedonaserverandnotusuallyaccessiblethroughafilesystemordirectory structure.Themainoptionsforhostingadatabasesis:onasingleserver,inadatabasecluster, orasacloudservice.Allofthesesystemsaredesignedtoabstracttheimplementationdetails. Aclienthassoftwarethatallowsittoconnectandcommunicatewiththedatabaseserverusing astandardizedprotocol.Therearemanydifferentuserinterfacesformanydatabases. DatabasescanbeaccessedremotelyorontheInternet. Learning Objectives • • • • • Identifyandunderstanddatabasenormalization. Reviewwhywenormalizeourdatamodels. Understandthestep-by-stepprocessofdatanormalizationandforms. Learnandapplyhowwenormalizearelationaldatamodel. Recognizethedrawbacksofnormalization. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 13 Database Queries Summary Aswecontinueourdiscussionofdatabasemanagement,wediveintotheissueofdatabase queries.Theabilitytomakeeffectivequeriesinalargedatabaseenablesustoharnessthe powerofbigdatasets.SQL(StructuredQueryLanguage)isalanguageandcontainsthe commandsweusetocreate,manage,andqueryrelationaldatabases.Asinalltechnologyand systemsapplications,thereareamultitudeofvendorswhoofferSQLvariations,butingeneral theyhaveacommonsetofdatatypesandcommands.SQLisportableacrossoperatingsystems andingeneral,portableamongvendors.Havingcoveredthecommonlyuseddatatypesin previouslessons,inthisnextsectionwewillcoververycommonlyusedqueries. Structured Query Language SQLisusedtoquery,insert,update,andmodifydata.UnlikeJava,VisualBasic,orC++,SQLis notacompleteprogramminglanguage;itisasub-languageofapproximately30statement types.Itisgenerallyembeddedinanotherlanguageortooltoenabledatabaseaccess.Afew definitionsweneedtobeawareofasweexploreSQLare: • Datadefinition:Operationstobuildtablesandviews(virtualtables) • Datamanipulation:INSERT,DELETE,UPDATEorretrieve(SELECT)data • Dataintegrity:Referentialintegrityandtransactionsenforceprimaryandforeignkeys • Accesscontrol:Securityformultipletypesofusers • Datasharing:Databaseaccessedbyconcurrentusers AfewissuestomakenoteofasyouworkwithSQListhatithasseveralinconsistencies.For example,NULLscanbeproblematicandwewillexplorethatlater.Inaddition,whenworking withSQLitisimportanttorecognizethatitoperationsondeclarativelanguage,notprocedural language.Thismeansthatyouwritethecommandinsuchawaythatdescribeswhatyouwant todo,notHOWyouwanttodoit.Itisleftuptotheapplicationtofigureitout. Variations among SQL Implementation BecausedifferentdatabasesuseSQL,therecanbevariationinhowSQLisimplemented.The variationsinclude: • Errorcodes • Datatypessupported(dates/times,currency,string/textvariations) • Whethercasematters(upper,lowercase) • Systemtables(thestructureofthedatabaseitself) • Programminginterface(novendorfollowsthestandard) • Reportandquerygenerationtools • Implementer-definedvariationswithinthestandard • Databaseinitialization,openingandconnection V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 14 Aswehavealreadylearned,adatatypedefineswhatkindofvalueacolumncancontain. However,becausethereisvariationacrossdatabases–wewillusethedatatypesforMySQL forthepurposeofthisdiscussion.MySQLhasthreemaindatatypes:numeric,strings(text), anddates/times.Seethefollowingfigures: CoreMySQLDataTypes-Numeric CoreMySQLDataTypes– Numeric NumericType Description INT Astandardinteger BIGINT Alargeinteger DECIMAL Afixed-pointnumber FLOAT Asingle-precision,floating-pointnumber DOUBLE Adouble-precision,floating-point number BIT Abitfield CoreMySQLDataTypes– Strings(Text) CoreMySQLDataTypes–Strings(Text) String Type Description CHAR Afixed-length,non-binarystring(character) VARCHAR Avariable-length,non-binarystring NCHAR Sameasabove +UnicodeSupport NVARCHAR Sameasabove +UnicodeSupport BINARY Afixed-length,binarystring VARBINARY Avariable-length,binarystring TINYBLOB AverysmallBLOB(binarylargeobject) BLOB AsmallBLOB TEXT Asmall,non-binarystring V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 7 8 15 CoreMySQLDataTypes– Dates/Times CoreMySQLDataTypes–Dates/Times Date /TimeType Description DATE Adatevaluein'CCYY-MM-DD' format TIME Atimevaluein'hh:mm:ss' format DATETIME Date/Timein'CCYY-MM-DDhh:mm:ss’format TIMESTAMP Timestampin'CCYY-MM-DDhh:mm:ss’format YEAR AyearvalueinCCYYorYYformat Creating Databases and Tables Togetstarted,wewillneedtoknowhowtocreatedatabasesandtables.WhileMySQLcanbe anintimidatingprogram,butonceyoumastersomeofthebasics,youwillbeabletowork effectivelywithlargedatasets. • TocreateaDatabase,weusetheCREATEDATABASEcommand 9 • Onceyouhavecreatedthedatabase,youwillnowapplytheUSEcommandtotellthe systemwhichdatabasetouse Onceyouhavecreatedadatabase,youwillwanttocreatetableswithinthelargerdatabase: • NewtablesaredeclaredusingtheCREATETABLEcommand • Wecanalsosetthenameanddatatypeofeachattribute • Whencreatingnewtables,wecanspecifyprimarykeysandforeignkeyrelationships • WecanalsodecidewhetherornotNULLoremptyvaluesareallowed Inserting Data into a new Database Onceyouhavecreatedanewdatabase,youarereadytoinsertdata.Thedatamodelwillacta guidetoloaddataintoanewdatabase.Ifthedatabasebuildswell,itmaymeanthatyouhave foundtherealbusinessrules.Or,ifyouhavesomeerrors,youmayhavetherealbusinessrules, butthedatamaybemessy.Finally,ifitbuildswithmanyerrors–thismaybethecasethatthe businessrulesarenotaccurateandareclosertowhattheywantorthinktheyhave,notwhat theyuse.Inmanycases,itisusefultogetsampledataandbrowseitduringtheprocessof buildingthemodel. SQL Select Queries SQLSELECTqueryisusedtofetchthedatafromadatabasetable,whichreturnsdatainthe formofaresulttable.SELECTreturnsasetofattributesinaquery.Inmostapplications,SELECT isthemostcommonlyuseddatamanipulationlanguagecommand.SELECTstatementsare constructedfromaseriesclausestogetrecordsfromoneormoretablesorviews. Clausesmustbeinorder;onlySELECTandFROMarerequired: • SELECTattributes/columns V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 16 • • • • • • • INTOnewtable FROMtableorview WHEREspecificrecordsorajoiniscreated GROUPBYgroupingconditions(attributes) HAVINGgroup-property(specificrecords) ORDERBYorderingcriterionASC|DESC DISTINCTreturndistinctvalues WildcardsinSQL Mostdatabaseimplementationsofferadditionalregularexpressions–wildcards.Awildcard charactercanbeusedtosubstituteforanyothercharacter(s)inastring.Regularexpressions canbeusedtofindrecords,whichmatchcomplexstringpatterns.Forinstance,MySQLhas: • [list]matchanysinglecharacterinlist,e.g.[a-f] • [^list]matchanysinglecharacternotinlist,e.g.[^h-m] Editing a Table Insomecasesyouwillbefacedwiththeneedtoeditatable.Inthiscaseyouwillusethe following: • INSERTisusedtoaddanewrecordtoatablethatcontainsspecificvaluesforasetof attributesinthattable • TheUPDATEkeywordisusedtomodifyaspecificvalueorsetofvaluesforasetof recordsinatable • DELETEisusedtoremoverecordsfromatablethatmeetaspecificcondition Learning Objectives • • • • • • • BecomemorefamiliarwithSQL. RecognizedifferentimplementationsofSQLhavedifferencesofwhichtobeaware. Reviewthedifferentdatatypes. Learnhowtocreatenewdatabasesandtables. UnderstandhowtouseaSELECTquery. Befamiliarwithwildcardsandwhentousethem. Reviewhowtoeditatable. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 17 Database Conditional, Grouping, and Joins Summary Inthenextsectionweexaminehowtodealwithdatabaseconditional,grouping,andjoins.As wegetfurtherintoSQL,wewillneedtorefineourapproachtomakeouractionsmore effective.Forexample,wewillneedtonarrowthesetofrecordsthatgetreturnedfroma query.Wewillalsoneedtomakestatisticalqueriesacrossdifferentgroupingsorrecords.In addition,wewillneedtosampleororderouroutputresults.Anotherchallengewillinclude integratingdatafromothersourceswithinourdatabase.Thefollowingreviewwillcoverthese challengesandothersaswecontinuetoworkwithSQL. Database Conditional Clauses Aconditionalclauseisapartofaquerythatrestrictsrowsmatchedbycertainconditions.You cannarrowSELECTstatementswithconditionalclausessuchasWHEREIN,ortheBETWEEN keyword.WHEREINstatementsareusedtoidentifyrecordsinatablewithanattribute matchingavaluefromaspecifiedsetofvalues.TheBETWEENkeywordsareusedtoidentify recordsthathavevaluesforaparticularattributethatfallwithinaspecifiedrange WHEREIN:WHEREattributeINisusedtoselectrowsthataresatisfiedbyasetofWHERE conditionsonthesameattribute.Example: SELECT* FROMOffices WHEREStateIN('CO','UT','TX'); isequivalentto: SELECT* FROMOffices WHEREState='CO'ORState='UT' ORState='TX' BETWEENKeyword:Selectrecordswheretheattributevalueisbetweentwonumbersusing BETWEEN,rangeisinclusiveandalsoworkswithtimeanddatedata. Null Values Nullvaluesaretreateddifferentlyfromothervalues;theyareusedasaplaceholderfor unknownorinapplicablevalues.Ifvaluesareemptyormissing,theyarestoredasNULL.Afew issuestobeawareofforNULLvalues: • NULLvaluesevaluatetoNOTTRUEinallcases • CheckforNULLSusingISandISNOT V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 18 • WhenaspecificattributemaycontainNULLormissingvalues,specialcaremustbe takenwhenusingconditionalclauses Grouping Data and Statistical Functions OnceyouareabitmorecomfortableworkingwithSQL,youcanstarttoexploresomeofthe statisticalfunctionsthatareincludedinmanyimplementationsofSQL.Thesefunctionscan operateonagroupofrecords.UsingtheGROUPBYclausewillreturnasinglevalueforeach groupofrecords.TofurtherrestricttheoutputoftheGROUPBYclausetoresultswithcertain conditions,usetheHAVINGkeywords(analogoustotheWHEREclause). AggregateStatisticalFunctionsinSQL Commonlyusedfunctionsinclude: Moreadvancedstatisticalfunctionscanbecreatedusingthebasicstatisticalfunctionsbuiltinto SQLsuchascalculatingtheweightedaverageorgettingthez-scorevaluesbycombining differentfunctions. Sorting and Sampling Data Youwillalsobefacedwiththeneedtosortandsamplethedata.Severalclauseswillhelpyou willthatincludingORDERBY,LIMIT,andRAND. ORDERBY:TheORDERBYclausespecifiesthattheresultsfromaqueryshouldbereturnedin ascendingordescendingorder LIMITthenumberofreturnedrecords:ALIMITclauserestrictsthenumberofrecordsthat wouldbereturnedtoasubset,whichcanbeconvenientforinspectionorefficiency V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 19 Randomlyselectandorderrecords:TheRAND()functioncanbeusedtogeneraterandom valuesintheoutputortorandomlysampleorrandomlyordertheresultsofaquery.For instance: Reordertheentiretable: SELECT* FROMtable ORDERBYRAND(); Randomlyselectasinglerecord: SELECT* FROMtable ORDERBYRAND() LIMIT1; Generatearandomnumberintheoutputresults: SELECTid,RAND() FROMtable; Creating New Tables and Aliases ASKeyword(Aliases):TheASkeywordcreatesanaliasforanattributeorresultofafunction thatisreturnedinaquery CREATETABLEAS:UseCREATETABLEwithAStocreateanewtableinthedatabaseusinga selectquery.Itmatchescolumnsanddatatypesbasedontheresultsintheselectstatement. ResultsfromaquerycanbeinsertedintoanewtableusingtheCREATETABLEwiththeAS keyword.Asseeninthefollowing: CREATETABLEnew_table AS( SELECTcolumn_name(s) FROMold_table); SELECTINTO:ResultsfromaquerycanbeinsertedintoanexistingtableusingaSELECTINTO clauseifthetablewiththeappropriatestructurealreadyexists.Taketheresultsofaselect statementandputtheminanexistingtableordatabase: SELECTcolumn_name(s) INTOnewtable[INexternaldb] FROMtable1; V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 20 Joining Multiple Tables Therelationaldatabasemodelallowsusjoinmultipletablestobuildnewandunanticipated relationships.Thecolumnsinajoinmustbeofmatchingtypesandalsomustrepresentthe sameconceptintwodifferenttables.Thiscanhelpustocontextualizeorintegrateatablein ourdatabasewithdatafromanexternalsource. Wewanttolearnhowtotakedatafromdifferenttablesandcombineittogether.Thismay includedatafromotherdatasourcesthatcomplementourown,suchasdemographic informationforazipcodeorpricestructureforshippingzonesforacarrier.Theprocessof mergingtwoseparatetablesiscalled“joining”.Joinsmaybedoneonanycolumnsintwo tables,aslongasthemergeoperationmakeslogicalsense.Seebelowforavisual representationofjoining: Joins-visually Twotableswithshared field/data SELECT * FROM tb1, tb2; SELECT * FROM tb1, tb2 WHERE tb1.bg = tb2.bg; SELECT grey, pink FROM tb1, tb2 WHERE tb1.bg = tb2.bg; ColumnsinaJOIN • Theydon’tneedtobekeys,thoughtheyusuallyare • Joincolumnsmusthavecompatibledatatypes • Joincolumnisusuallykeycolumn:Eitherprimaryorforeign • NULLswillneverjoin 36 Types of Joins and Views Joinfrom3Tables:Joiningthreetablestogetherjustinvolvesoneadditionaljoinbetweentwo alreadyjoinedtablesandathirdtable. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 21 JoinTypes Differenttypesofjoinscanbeusedtomergetwotablestogetherthatalwaysincludeeveryrow inthelefttable,righttable,orinbothtables.ThefollowingaredifferenttypesofJOIN: • INNERJOIN:returnsonlytherecordswithmatchingkeys(joinscommoncolumnvalues) • LEFTJOIN:returnsallrowsfromLEFT(first)table,whetherornottheymatcharecordin thesecondtable • RIGHTJOIN:returnsallrowsfromRIGHT(second)table,whetherornottheymatcha recordinthefirsttable • OUTERJOIN:Returnsallrowsfrombothtables,whetherornottheymatch(Microsoft SQL,notMySQL) • InMySQL,JOINandINNERJOINareequivalent Views Viewsarevirtualtablesthatdonotchangetheunderlyingdatabutcanbehelpfultogenerate reportsandsimplifycomplicatedqueries.Theyarevirtualtablesthatpresentdataina denormalizedformtousers.Theydonocreateseparatecopiesofthedata(theyreferencethe dataintheunderlyingtables).Thedatabasestoresadefinitionofaviewandthedatais updatedeachtimetheVIEWisinvoked. ThereareseveraladvantagestoVIEWS.Userqueriesaresimpleronviewsconstructedfor them.Theyoffersalayerofsecuritythatcanrestrictaccesstodatainviewsforusers.Theyalso providegreaterindependence,meaningthattheuserorprogramisnotaffectedbysmall changesinunderlyingtables. Learning Objectives • • • • • • • • • • LearnhowtoworkwithSELECTforconditionalclauses. RecognizetheroleanduseofNULLvalues. ReviewhowtogroupdatawiththeGROUPBYclause. IntroducetheexistenceofstatisticalfunctionsinallSQLfunctions. Recognizehowtoapplyaggregatestatisticalfunctions. ReviewsortingandsamplingtechniquessuchasORDERBY,LIMITandRAND. LearnhowtocreatenewtablesandaliasesusingtheASkeyword. Becomingfamiliarwithjoiningmultipletables. RecognizethetypesofJOINs. IdentifywhenandhowtouseVIEWs. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 22 Introduction to Machine Learning Summary Inthislessonweexploremachinelearning.Thisincludesidentifyingwhenweneedmachine learninginsteadofothertechniquessuchasregression.Webreakdownthedifferentclassesof machinelearningalgorithms.Inaddition,weidentifyhowtousemachine-learningapproaches tomakeinferencesaboutnewdata. Review of Regression Linearregressionusesthevalueofoneormorevariablestomakeapredictionaboutthevalue ofanoutcomevariable.Inputvariablesarecalledindependentvariablesandtheoutput variableisknownasthedependentvariable. • Linearregressionoutputincludescoefficientsforeachindependentvariable. o Thisisameasureofhowmuchanindependentvariablecontributestothe predictionofthedependentvariable. o Theoutputalsoincludesmetricstobeabletoassesshowthemodelfitsthe data.Thebetterfitofthemodel,thebetteryouareabletomakeaccurate predictionsaboutnewdata. • Usingcoefficientscalculatedfromhistoricdata,aregressionmodelcanbeusedtomake predictionsaboutthevalueoftheoutcomevariablefornewrecords. Overview of Machine Learning Algorithms Machinelearningalgorithmsareprimarilyusetomakepredictionsorlearnaboutnew, unlabeleddata.Thereareseveralclassesofalgorithms: • Classification:assigningrecordstopre-defineddiscretegroups • Clustering:splittingrecordsintodiscretegroupsbasedonsimilarity;groupsarenot knownapriori • Regression:predictingvalueofacontinuousordiscretevariable • Associatelearning:observingwhichvaluesappeartogetherfrequently Supervised vs. Unsupervised Machine Learning Supervisedlearningusesoutcomevariables,knownaslabels,foreachrecordtoidentify patternsintheinputvariablesorfeaturesrelatedtotheoutcomevariable. • Correctanswer,labelisknowninthetrainingdata • Labelisappliedbyapersonoralreadyexists • Labeleddataareusedtotrainanalgorithmusingfeedback • Applyortestthetrainedmodelonnew,unseendatatopredictthelabel V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 23 Supervisedlearningworkflow Supervisedlearningworkflow LearningFlow Training Training RawData Labels RawData Labels Algorithm (Training) Algorithm (Training) Model Model EvaluaBon MakingPredicBons MakingPredicBons Unlabeled Data Unlabeled Data Model Model PredicBon PredicBon Unsupervisedlearningworkflow LearningFlow EvaluaBon 16 Inunsupervisedlearning,theoutcomevariablevaluesareunknown,therefore RawData relationshipsamongtheinputvariablesareusedtoidentifypatternsofclustersof records. 16 • Findspreviouslyunknownpatternsinthedatawithoutlabelsorguidance Algorithm • Notraining/testing/validatingprocessbecausecorrectanswerisunknown Model Model Quality Machinelearningmodelsshouldbetrainedonanunbiasedsetofdatathatis Produc<on representativeofthevarianceintheoveralldataset.Biasquantifiesthelackof abilityofamodeltocaptureunderlyingtrendinthedata.Morecomplexmodels decreasebiasbuttendtoincreasevariance.Variancequantifiesamodel’ssensitivitytosmall changesintheunderlyingdataset. • Ideallywantlowbiasandlowvariance,butthereisatradeoffbetweenthetwo quantities • Ifthereisabiasinthetrainingdataoriftoomanyfeaturesareincludedinamodel,the modelisatriskofbeingoverfit.Inoverfitmodels,thecoefficients,knownas parameters,willnotbegeneralizableenoughtomakegoodpredictionsfornewrecords. • Alargeandrepresentativesampleofthelabeleddatashouldbeusedtotrainthe model,theremainderisusedfortesting. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 24 Overfitting vs underfitting • • Underfitting–modelistoosimple,highbiasandlowvariance Overfitting–modelistoocomplex,lowbiasandhighvariance o Overfittingisamorecommonpitfall o Temptingtomakemodelsfitbetterbyaddingmorefeatures o Resultsinamodelthatisincapableofgeneralizingbeyondthetrainingdata Learning Objectives • • • • • Beintroducedtomachinelearning Becomefamiliarwithdifferenttypesofmachinelearningalgorithms Beabletodifferentiatesupervisedandunsupervisedlearningandtheirprocesses Recognizemodelqualityandthetradeoffsbetweenbiasandvariance Learnhowtoidentifywhenamodelisoverorunderfit V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 25 Machine Learning Algorithms Summary Inthislessonwearegoingtodivedeeperintomachinelearningalgorithms.Eachmodelhas differentpropertiesandisbestfordifferenttypesoftasks.Wereviewhowtocomparethem withperformancemetrics.Weneedtobeabletogrouprecordstogetherwithoutlabelsto informpredictionusingunsupervisedclassification.Inaddition,wereviewcapabilityto confidentlyreducethenumberoffeaturesincludedinananalysiswithoutlosinginformation. Thelessonalsointroduceshowtocomparepredictoraccuracyandtestforsensitivityand specificity. Dimensionality reduction Dimensionalityreductionisatermforreducingfeaturesincludedinanalysis.Itisoftenneeded foranalysiswithmanyfeatures.Tryingtoreducedimensionalityrandomlyormanuallyleadsto poorresults • Resultsneedtobeinterpretedbyhumans,shouldbetractable • Increasingthenumberoffeaturesincludedincreasestherequiredsamplesize • Featuresshouldnotbeincludedordiscardedfromanalysisbasedoninstinct o Dimensionalityreductiontechniquesshouldbeemployed,suchasprincipal componentanalysis. • Summarystatisticsareameansofdimensionalityreduction Principal component analysis (PCA) PCAisamathematicalapproachtoreducedimensionalityforanalysisorvisualization.It exploitscorrelationstotransformthedatasuchthatthefirstfewdimensionorfeaturescontain amajorityoftheinformationofvarianceinthedataset.PCAdetermineswhichvariablesare mostinformativebasedonthedistributionofdataandcalculatesthemostinformative combinationsofexistingvariableswithinthedataset.PCAworkswellfordatasetswithhigh dimensionality. • Noinformationislost,firstfewcomponentsholdmuchoftheinformation • Samepremiseaslinearregressionexceptwithoutadependentvariable o Linearregressionsolutionisthefirstprincipalcomponent o Disregardingtheinformationdescribingtheprincipalcomponent,PCAcalculates thesecondmostinformativecomponent,thenthethird,andsoon • Linearcombinationsformasetofvariablesthatcanbeusedtoviewthedata–new axes • Componentsarerankedbyimportance,soallbutthefirstfewcanbediscarded,leaving onlythemostimportantinformationwithveryfewcomponents • Thecoefficientsinthetablegivetheproportionofeachoftheoriginalvariablesthat wentintoeachcomponent V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 26 • • Relativesigns+/-indicatethattwovariablesarepositivelynegativelycorrelatedinthat particularcomponent Thecomponentsaredifficulttointerpretusingonlythecoefficientvalues,plottingoften improvesunderstanding PC1=(a*var1+b*var2+c*var3+…) PC2=(d*var1+e*var2+f*var3+…) PC3=(g*var1+h*var2+i*var3+…) Clustering Anotherwayofthinkingaboutdimensionalityreductionishowcloseeachpointistoother points.Theideaistoseparatedatapointsintoanumberofclustersthathavelessdistance betweenthepointsinternallythantootherclusters.Clusteringcanbehelpfultoidentifygroups ofrecordsthathavesimilarcharacteristicstooneanother.Whendataisunlabeled,clustering canbeusedtogrouprecordstogetherfordeeperinspection.Upondeeperinspectionofthe recordsineachcluster,userscanunderstandthepatternsthatleadtorecordsbeinggrouped together,andalsoidentifyreasonsforrecordsbeinggroupedseparately. K-means clustering k-meansclusteringstartswithselectingthenumberofclusters,k.kcluster-centersareplaced randomlyinthedataspaceandthenthefollowingstagesareperformedrepeatedlyuntil convergence.K-meansdoesnotdeterminetheappropriatenumberofclusters,thisissetbythe userbasedonintuitionorpreviousknowledgeofthedataThealgorithmcanterminatewith multiplesolutionsdependingoninitialrandompositionsofcluster-centersandsomesolutions arebetterthanothers. • Datapointsareclassifiedbythecentertowhichtheyarenearest • Thecentroidofeachclusteriscalculated • Centersareupdatedtothecentroidlocation Classifications • • • • ClusteringandPCAallowuserstoseepatternsinthedata,whichisthebestthatcanbe donebecausetherearenolabelstoguidetheanalysis Withsupervisedlearning,thelabelisincludedinthelearningprocess: o Unsupervised:whatfeaturesaremostimportantorinteresting? o Supervised:whatfeaturesaremostinformativeaboutthedifferencesbetween thesegroups? Classificationmethods:eachrecordfallsintosomecategoryorclass,predictthe categoryofanewrecordbasedonvaluesofotherfeaturesintherecord Regressionmethods:onevariabledependsonsomeorallofothers,predictthevalueof thedependentvariablebasedonthevaluesoftheindependentvariables V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 27 Classification Trees Classificationtreessplitdatatofindoptimalvaluesforfeatures,usedtosplitdatabyclass.Tree diagramsshowtheclassmakeupofeachnode,andtherelativenumberofdatapointsthat reacheachnode • Treepruning o Treepruningremovesrulesassociatedwithoverfittingfromthetree o Thenewtreemissesafewpointsclassifiedcorrectly,butcontainsonlymeaningful rules,moregeneralizabletonewdata Naïve Bayes classifier • • • TheNaïveBayesalgorithmconsidersthevalueofeachfeatureindependently,foreach record,andcomputestheprobabilitythatarecordfallsintoeachcategory Next,theprobabilitiesassociatedwitheachfeaturearecombinedforeachclass accordingtoBayes'ruletodeterminethemostlikelycategoryforeachnewrecord Almostcompletelyimmunetooverfitting-Individualpointshaveminimalinfluence; Veryfewassumptionsaremadeaboutthedata Random forest Randomforestisanensembleclassifierthatusesmultipledifferentclassificationtrees.Trees aregeneratedusingrandomsamplesofrecordsintheoriginaltrainingset.Accuracyand informationaboutvariableimportanceisprovidedwiththeresult. • Nopruningnecessary • Treescanbegrownuntileachnodecontainsveryfewobservations • Betterpredictionthanclassification • Noparametertuningnecessary V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 28 Comparing predictor accuracy Cross - validation • • • Modelsshouldbegoodatmakingclassificationsofunlabeleddata,notdescribingdata thatisalreadyclassified. Randomlydividedataintoatrainingsetandatestset o Hidetestsetwhilebuildingthetree o Hidetrainingsetwhilecalculatingaccuracy o Computedaccuracyrepresentsaccuracyonunseendata Techniquesareavailabletodothismultipletimes,ensuringeachrecordisinthetestset exactlyonce,e.g.k-folds Comparing models • • Severalstandardmeasuresofperformanceexist,canrunmultiplemodelsandcompare metrics: o Accuracy o Precision o Recall o Andmore Applicationdriveswhichperformancemetricsaremostimportantforagiventask Sensitivity and specificity Sensitivityandspecificityarestatisticalmeasuresoftheperformanceofaclassificationtest. Sensitivitymeasurestheproportionofpositivesareidentified.Specificitymeasuresthe proportionofnegativesthatareidentified.Seeexamplebelow: V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 29 The ROC Curve • • • • TheReceiverOperatingCharacteristic(ROC)curveplotsthetruepositiverate (Sensitivity)versusthefalsepositiverate(100-Specificity)fordifferentcut-offpoints Eachpointonthecurverepresentsapairofsensitivity/specificityvaluescorresponding toaparticulardecisionthreshold Atestwithperfectdiscrimination(nooverlapinthetwodistributions)hasanROCcurve thatpassesthroughtheupperleftcorner(100%sensitivity,100%specificity) TheclosertheROCcurveistotheupperleftcorner,thehighertheoverallaccuracyof thetest Learning Objectives • • • • • • Befurtherintroducedtomachinelearningalgorithmsandhowtoworkwiththem Becomefamiliarwithdimensionalityreductionandwhenandhowtouseit Recognizewhentouseclusteringasanapproachtodimensionalityreduction Reviewdifferentclassificationmethodssuchasclassificationtressandrandomforest Learnhowtocomparepredictoraccuracy Becomefamiliarwithsensitivityandspecificityasindicatorsofabinaryclassification test V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 30 Supply Chain Systems - ERP Summary Inthisnextsegment,weexploresupplychainITsystems.Becausesupplychainsareessentially madeupofthreeflows:information,material,andmoney–ITsystemssupporttheinformation flow.Forexample,inasupermarket,theyhavetodealwithdifferenttypesofsupplychaindata suchassupplierinventory,facilitymanagementandpayroll,sales,andexpiredandobsolete inventory.Therearemanydailytransactionsinasupermarketthatneedtobecapturedand ensuredforconsistencyandtimeliness.Thatinformationneedstobesomehowtranslatedinto usableinformationforbusinessdecisions,andthentheseobjectivesneedtobeefficiently achieved.Theamountofinformationfortransactionsperweekinasinglesupermarketcan numberinthemillions.Thisisforasinglestore. Onanenterpriselevel,companiesneedsystemsthathelpthemmanageandorganizethis informationforuse.Whilesupplychainsarealwaysportrayedasneatandlinearsystems,the realityismuchdifferent,aswehavelearnedoverthepreviouscourses.Flowsmoveupand downthechainandthroughmanypartnersuntiltheyreachtheirfinaldestination.Supply chainsneedITsystemsbecausewhileteamsmaysitindifferentfunctionalunitstheyfrequently needtoshareinformation.Inaddition,manydifferentfirmsinteractinthesupplychain,they needsystemstocarrythatinformationbetweenthem,thishelpsde-silothesupplychain. Thereneedstobecoordinationacrossfunctions,whichistheessenceofsupplychain managementandcanbefacilitatedwithsystemslikeEnterpriseResourcePlanning(ERP). Supply Chain IT Systems SupplychainsneedITsystemsbecausetheyarelarge,complexandinvolvemanyplayers.They oftenbecomeintertwinedandindividualactorsimpactothers.Decision-makingisbasedon commondataandinteractionwithotherfunctionsinafirm.AndsupplychainsneedITsystems becausesupplychainsrequirecommunicationforsomanyinteractionsB2B,B2C,M2M,etc. (B2B=businesstobusiness,B2C=BusinesstoConsumer,M2M=machinetomachine) EnterpriseResourcePlanning(ERP)systemsserveasageneralledgerandcentraldatabasefor allfirmactivity.ThenextareSupplyChainPlanningSystems.Thesesystemsareprimarilyfor productionplanningandscheduling,demandplanningandproductlifecyclemanagement.The lastareforSupplyChainExecution;whicharetransportationandwarehousemanagement systemsandmanufacturingsystems.ThefirstwewilltackleareEnterpriseResourcePlanning systems. Enterprise Resource Planning InthefollowingsectionwecoverwhyfirmsuseERPs;thecorefunctionsofERPs;dataneeded; communicationsofthesystems;andstrategicbenefitsofanERP.MostfirmshaveanERP becausemanyfunctionsinafirmsuchassales,inventory,production,finance,distribution,and V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 31 humanresourceshavesiloeddatabases.WithacentralizedERP,thesedatabasescanmore easilybeexchangedandshared. BenefitsofERPallowenterprisestoorganizeprocessesanddatastructure,integrate informationintounifiedrepository,makedataavailableformanyusers,eliminateredundant systemsanddata,reducenon-valueaddedtasks,standardizeprocessdesigns,aswellasbe moreflexible.TherearesignificantdrawbackstousingERP,theseinclude:significant implementationtimeandmaintenancethatcomeatacost,dataerrorsripplethroughsystems, competitiveadvantagecanbedampened,firmrelianceonasinglevendor,shortageof personnelwithtechnicalknowledgeofsystem,andhighimpactofdowntimeofsaidsystem. ERP Core Functions MostERPSystemssharethesamecorefunctions.Theytietogetherandautomateenterprisewidebasicbusinessprocesses: CustomerManagementisthefacetoconsumersandservesthefollowingfunctions: • enablesorderentry,orderpromising,openorderstatus • allowsmarketingtosetpricingschemes,promotions,anddiscounts • providesreal-timeprofitabilityanalysis,and • permitsorderconfiguration,customerdeliveryschedules,customerreturns,tax management,currencyconversion,etc. Manufacturingisthefacetoproductionandservesthefollowingfunctions: • includesMRPprocessing,manufacturingorderrelease,WIPmanagement,cost reporting,shopfloorcontroletc., • providesrealtimelinkageofdemandtosupplymanagementenablingrealtime Available-to-Promise(ATP)&Capable-to-Promise(CTP),and • servesasprimaryinterfaceto“bolt-on”advancedplanningandschedulingoptimization modules. Procurementisthefacetosuppliersandservesthefollowingfunctions: • integratesprocurementwithsuppliermanagement, • facilitatespurchaseorderprocessing,deliveryscheduling,openordertracking, receiving,inspection,andsupplierperformancereporting,and • createsrequestsforquotation(RFQ) • managesnegotiationandpricingcapabilities. Logisticsisthefacetointernalandexternalsupplychainandservesthefollowingfunctions: • runstheinternalsupplychainforenterprise, • providesconnectivitytoremotetradingpartners(3PLs,carriers,etc.), • handlesdistributionchannelconfiguration,warehouseactivitymanagement,channel replenishment,planning,distributionordermanagement,etc.,and • servesasprimaryinterfaceto“bolt-on”warehouseandtransportationmanagement systems(WMSandTMS). ProductDataisthefacetoallmaterialandservesthefollowingfunctions: • describesproductsenterprisemakesand/ordistributes, V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 32 containsproprietarydataoncosts,sources,engineeringdetails,dimensions,weight, packaging,etc., • interfaceswithinventory,manufacturing,andproductlifecyclemanagement,and • sometimesincludedinpartnercollaborationsinordertocompresstimetomarketof newproducts. FinanceisthefacetotheCFOandservesthefollowingfunctions: • strongsuitofmostERPs(butalsoadoubleedgedsword!), • providesreal-timereportingofalltransactionsresultingfrominventorymovement, accountsreceivable,accountspayable,taxes,foreigncurrencyconversions,andany otherjournalentries,and • supportsdetailedreportingandbudgetingcapabilities. • ERP Data TherearethreetypesofERPData: Organizationdata:representsandcapturesthestructureofanenterprise. Masterdata:representsentities(customers,vendors)withprocesses.Itisthemostcommonly used.Butbecausespecificprocessesusematerialsdifferentlyandspecificdataneedsdifferby processes–thisaddstocomplexityofmasterdataneeds.Materialtypescanbeindifferent statesandcanbegroupeddifferentlybasedonfirmneeds. Transactiondata:reflectstheoutcomeofexecutingprocesssteps.Itcomesinorganizational, masterandsituationaldata.Transactiondocumentsincludepurchaseorders,invoices,etc. ERP Communication Business-to-Business(B2B):Commercetransactionsbetweenmanufacturers,wholesalers, retailers.Eachbusinessrepresentsalinkinthesupplychain. Business-to-Consumer(B2C):Saletransactionsbetweenfirmsandend-consumers.Thevolume ofB2BtransactionsismuchgreaterthanB2C. AcceleratingandvalidatingB2BandB2Ctransactions:ForB2Bthisisachievedthrough ElectronicDataInterchange(EDI).ForB2Cthisisachievedthroughawebsiteandemail. ElectronicDataInterchange(EDI):“Thecomputer-to-computerinterchangeofstrictly formattedmessagesthatrepresentdocumentsotherthanmonetaryinstruments.”Thereisno humaninterventionintheprocess. ERPsystemscan“communicate”viaEDI,sharingnearreal-timeinformation.Seefigurebelow. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 33 ERPCommunica-on § B2BEDIexample: PurchaseOrder OrderConfirma-on SAP Data Format OrderCancella-on BusinessX Gateway ETA OrderCancella-on BusinessY Gateway Oracle Data Format ASN OrderReceipt § ERPsystemscan“communicate”viaEDI,sharingnearreal--meinforma-on. The§ Value of ERP for SCM ThedataisusuallytranslatedandvalidatedtobeimportedintoanERPsystem. TherearethreeimportantvaluesofERPforsupplychainmanagement:reductionofthe § AnyinfofilecanbesharedgivenappropriateERPfieldstocaptureanddisplayits bullwhipeffect,enablingwidespreadanalytics,andextendingtheenterprise. content. § Whatotherinfowouldbusinesseswanttoshare? Reducing the Impact of Bullwhip Effect MIT Center for OneofthekeyvaluesofanERPsystemisreducingorpreventingtheBullwhipEffect.The 24 Transportation & Logistics AdaptedfromOmarElwakil(2016) BullwhipEffectisphenomenonwhereinformationdistortionleadstoincreasingorder fluctuationsintheupstreamsupplychain(forecast-drivensupplychains).Itisdrivenbyseveral behavioralcauseslikeoverreactiontobacklogsandlackoftransparency.Therearealsomany operationalerrorssuchasforecastingerrors,lead-timevariability,andpromotions. ERPcanreducetheimpactoftheBullwhipEffectbyextendingvisibilitydownstreamto customerdemandandupstreamtoparticipantsenablingcollaborationandinformation sharing.Italsofacilitatespointofsalecapturingandhelpsreducebatchsizeanddemand variabilitythroughsmallerandmorefrequentorders. Enabling Analytics ERPsystemsplayakeyroleinenablinganalytics.Theyareprimarilyretrospective,serveasthe ledgerofthefirm,andprovidetheCFOwithfinancialsnapshots.Theyalsoenableotherforms ofanalyticsforBusinessIntelligence(BI):whichtransformsrawdataintomeaningful informationtoimprovebusinessdecision-making.Thesecanbedescriptive,predictive,and prescriptive. Extending Enterprise WhileERPsystemsareprimarilyusedinintra-firmprocessmanagementtoconnectvarious departmentsandprovideaccesstodata,theyalsoserveanextendingfunctionforbetter connectionwithpartners.ERPsserveavalueinconnectingEndtoEndSupplyChainswith betterconnectionsacrossSCparticipants,providingsharedunderstanding,reducing coordinationandmonitoringcosts,andrespondingquicklytomarketfeedback. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 34 Learning Objectives • • • • • • IntroductiontosupplychainITsystems,theirvalue,application,andconstraints. ReviewERP,itssetupfunctionality,andapplications. RecognizecorefunctionsofERP. BefamiliarwithdatahouseinERPsystems. Reviewhowthatdataisusedtofacilitatecommunication. UnderstandsomeofthevalueofERPsystemsforsupplychains. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 35 Supply Chain Systems - Supply Chain Modules Summary InthisnextsegmentwereviewdifferentsupplychainmodulesasasubsetofsupplychainIT systems.TounderstandwherewearenowwithsupplychainITsystemsweneedtoreviewthe evolutionofsupplychaintools.Wejourneyfromthe1960-70’swiththeBillofMaterials Processor,mainframesbaseddatabasesystems,andmaterialrequirementsplanning(MRP)to the1980’swiththesecondcomingofMRPthataddedfinanceandmarketing,Just-In-Time manufacturingmethodology,expansiontootherfunctions,andtheprecursortotheERP systemsoftoday.Inthe1990s,mostMRPsystemswereabsorbedintoERPsuites;therewas theintroductionofAdvancePlanningSystems(APS),andwideradoptionofERPsystems.Inthe 2000s,manyofthesesystemsadoptedweb-basedinterfaces,improvedcommunication,and adoptedsharedorcloudbasedsolutions.Therewasalsoamajorconsolidationofsupplychain softwarevendorsandexpansionofERPsystemstoincludeSCM. Nowasweexplorehowtofurtherextendtheenterpriseanditsabilitytoadequatelymanage itsinformationonitsownandtogetherwithothercompanies,manyfirmsuseaseriesofIT modules.ThesesystemsaresometimesapartofERP,maybestandaloneapplications,orcan bepartofalargersupplychainecosystem.Wewillreviewtwomainfunctionalitiesincluding AdvancePlanningSystems(APS)andExecutionApplications.AdvancedPlanningSystems(APS) arelongrange,optimizationbaseddecisionsupporttoolswhileexecutionapplicationsinclude WarehouseManagementSystems(WMS),TransportationManagementSystems(TMS),and ManufacturingExecutionSystems(MES). Planning vs. Execution Althoughplanningmodulesseektoenablefutureplanningandenablingefficientprocesses, thereisoftenagapbetweentheplanningandexecutiontasks.Thefigurebelowillustratesthis Planning vs. Execution gap. Consists of a continuum of tasks, but . . . ROA Impact Supply Chain Strategy/ Network Design Execution Tasks Tactical Transportation Transportation Modeling Shipment Procurement Consolidation & Carrier Selection Planning Tasks Fleet Routing/ Scheduling . . . there is a gap! MIT Center for Transportation & Logistics V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 36 Questions, Approaches, and Technologies Change based on timeframe Questionscanbestrategicsuchas:“WhatcarriersshouldIpartnerwithandhow?”“How shouldIflowmyproducts?”ortheycanbetacticalsuchas:“HowcanIquicklysecureratesfor anewDC/plant/lane?”“Whatlanesarehavingperformanceproblems?”oroperational: “WhichcarriershouldItenderthisloadto?”“HowcanIcollaborativelysourcethisweeks’ loads?” Thetimeframealsodrivestheapproach.Forinstanceinthestrategicface–acompanywillbe establishingaplanandstrategy,haveeventbasedenablementandcompletenon-routine analysis.Inanoperationaltimeframetheywillbeexecutingthestrategicplan,operateon transactionbasedrulesandprocesses,andhaveautomatedactions. Andtechnologiesalsoalignwithtimeframes.Forinstance,strategictimelinewillallowfor analysisenginestoolslikeoptimization,simulationanddataanalysisandcommunicationvia theweb,fileexchangeandremoteaccess.Thetacticaltimelineallowsforthesameanalysis andcommunicationtechnologieswhiletheoperationaltimelineallowsforcommunicationbut alsoworkflowsoftwaresuchascompliancetracking,rules,andtransactionprocessing. Advanced Planning Systems Wenowtakeacloserlookatadvancedplanningsystemsthatareprimarilyusedasdecision supportsystems.Theytypicallyincludefunctionalityfornetworkdesign,demandplanning, productionplanning,productionscheduling,distributionplanning,andtransportationplanning. AdvancedPlanningSystemsutilizelargescalemixedintegerlinearprograms(MILPs)and sometimessimulation. Planning Horizons AdvancePlanningSystemshelpwithplanninghorizons.Thefollowingprovidearoughguideline buteachfirmdiffersanditisuniquetospecificindustries.: • 3monthsout–MasterProductionSchedule(MRP,DRP) o <4weeksout-FrozenMPS o 5to8weeksout–SlushMPS–somechangesallowed(+-10%) o >8weeksout–WaterMPS–morechangesareallowed(+-30%) • 3-18monthsout–AggregatedPlanning • >18monthsout–LongRangePlanning–NetworkDesign,etc. Flow Inputs(fromERPorothersystems):Currentcosts,manufacturingandstoragecapacities, consensusforecast,salesorders,productionstatus,purchaseorders,andinventorypolicy recommendations,etc. DecisionProcess:Largescaleoptimization(MILP)acrossmultiplefacilitiesandtimehorizonsin asingleplanningrun;unconstrained,constrained,andoptimalplans V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 37 Outputs:Demandforecastandplanformeetingdemand;afeasibleproductionplanforfuture periodstoincludeallocationofproductiontoplants;allocationoforderstosuppliers; identificationofbottlenecks;RootCauseAnalysis. Transportation Management Systems (TMS) TMSissoftwarethatfacilitatesprocurementoftransportationservices,short-termplanning andoptimizationoftransportationactivities,assets,andresources,andexecutionof transportationplans(Gonzalez2009).Itisoftengeographicandmodespecific.Thecore functionsofTMSare:transportationprocurement;modeandcarrierselection;carrier communication;routingguidegenerationandmaintenance;fleetmanagement;audit, payment,andclaims;appointmentscheduling;yardmanagement;androuteplanning. Transportation Execution TheTMSservesastheinterfacetothecarrierswhileconnectingtheOrderManagement System(OMS),PaymentSystems,andtheERP.Itsmainobjectiveisto:moveproductsfrom initialorigintofinaldestinationcosteffectivelywhilemeetingthelevelofservicestandardsand executingtheplanusingtheprocuredcarriers.Thisisbrokendowninphasesbelow: PLAN:CreateShipmentsfromOrders EXECUTE:SelectandtendertoCarriers MONITOR:VisibilityofthestatusofShipments RECONCILE:AuditinvoicesandpayforTransportation TherearemanyconsiderationstobemadeinTMSsuchas: § Howdoordersdrop?BatchedvsContinuous? § Howmuchtimeisallowedbetweendropandmust-ship?Weeks?Days?Hours? Minutes? § Whatpercentageoforderschangeafterrelease? § Howdotheychange?Quantity?Mix?Destinations?Timing? § Whatisthelengthofhaul? § Howmanyordersare“inplay”atanytime? Therearealsokeydecisionslikecarrierselectionandloadbuilding. TMS Carrier Communication & Selection UsefulEDITransactionSets • 204–MotorCarrierLoadTender:Usedbyshipperstotenderanofferforashipmentto afulltruckloadmotorcarrier.Itmaybeusedforcreating,updatingorreplacing,or cancelingashipment. • 990-ResponsetoaLoadTender:Usedbymotorcarrierstoindicatewhetheritwillpick upashipmentofferedbytheshipper • 214-TransportationCarrierShipmentStatusMessage:Usedbycarrierstoprovide shippersandconsigneeswiththestatusoftheirshipments. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 38 Carrier Selection Exampleofcarrierselection Carrier C Continuous Move Primary Carrier A Spot Carrier A Carrier E Spot Carrier B Carrier B Carrier B Carrier B Types of Capacity Dedicated Primary - Contracted Carrier Dedicated Fleet Continuous Move Spot Carrier Carrier D Spot 15 MIT Center for Transportation & Logistics LinkingApproaches Approaches Must Be Linked Tier III Spot execution – highly variable Tier II How do I select each carrier? Increased flexibility in execution Dynamic Tier I Uses strategic routing guide Flexible Assmt Dynamic Carrier Selection Dynamic Pricing in Private Exchange III IV I Static Strategic Lane Assmt Contract II Strategic Lane Assmt w/ Tier Pricing Dynamic How do I price each load? MIT Center for Transportation & Logistics 26 Warehouse Management Systems & Automation WMSisasoftwaresystemthatfacilitatesallaspectsofoperationswithinawarehouseor distributioncenterandintegrateswithothersystems.Itisnotthesameasinventory managementsystems;WMScomplementsIMS.ExamplesofthebenefitsofaWMSinclude: real-timestockvisibilityandtraceability,improvedlaborproductivity,andimprovedcustomer V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 39 service.Someofthesebenefitsarecloselytiedtoautomationofmaterialhandlingand paperlessdeviceinterfaces. ExamplesofWarehouseAutomationinclude: Automaticidentificationtechnologies:Barcodesandbarcodescanners,radiofrequencytags (RFID)andantennae,smartcardsandmagneticstripes,visionsystems. Automaticcommunicationtechnologies:Radiofrequencydatacommunications,synthesized voice,virtualdisplays,picktolight/voicesystems Automatedmaterialhandlingtechnologies:Carousels,conveyors/robotics,flowracks,AS/AR Systems WMS Software Components OrderProcessing • Orderchecking&batching • Allocation • Auto-replenishment Receiving • ASNplanning • Inboundtracking • Deliveryappointmentscheduling • POverification • Returnsprocessing Put-Away • Palletizing • Zoningandslotting • Random/directedputaway • Routingforputaway&replenishment Picking • Batch/Wave/Zone/Directedpicking • Carton/palletselect • Assembly/kitting • Pick-to-light/voice Shipping • Palletsequencing&Loadplanning • Palletlayering • Trailermanagement LaborManagement • Individual/teamperformancemgmt • Laborscheduling • Timestandards EquipmentSupport • Interfacetoautomatedequipment • Equipmentmaintenance V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 40 Manufacturing Execution Systems MESisasoftwaresystemthatmanagesandmonitorsallwork-in-process(WIP)inthe productionprocess.ItisintegratedwiththeERPtomanagetheexecutionofreleaseof productionorderstofinishedgoodsdelivery,triggersupplychainreplenishments,andenhance producttraceabilitythroughmanufacturing. ThefunctionalityofaMESinclude: • Machinescheduling • Processmanagement • Documentcontrol • Labormanagement • Inventorymanagement • Product(WIP)tracking • Performanceanalysis • Labormanagement • Qualitymanagement • Productionreporting Learning Objectives • • • • • Becomefamiliarwithsystemsthatarecommoninsupplychainsthatextendthe enterprise. DifferentiatebetweenAdvancedPlanningSystemsandExecutionSystems. Recognizethegapsinplanningvs.executionandthetimeframesembeddedinboth. ReviewAdvancedPlanningSystems,theiruseandapplication. BecomefamiliarwiththemainexecutionsystemsinSCsuchasTMS,WMS,andMES. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 41 Supply Chain Systems - Software Selection & Implementation Summary Aswehavejourneyedthroughvarioustypesofsupplychainsystems,wewillnowcoverthe processofsoftwareselectionandimplementation.Asfirmsareintheprocessofselectingtheir supplychainsystemstheywillneedtobeawareofvariousfactorsthatwillguidetheirdecision ofwhethertoselectthesystemornot.Sowenowdiscusstheprocessofsoftwareselection andevaluationcriteria.Whileselectingtheappropriatesoftwaresystemcanbechallenging, implementationisfarmoredifficult.Theprocessofimplementationcanbelongandcostly, usingadditionalresources.Inthislesson,wecovergeneralguidelinesonwhattobeprepared forwhenimplementingsoftwaresystems. Architecture Evolution of Architecture Tounderstandwheresupplychainsystemsstandnow,itishelpfultounderstandtheevolution ofthearchitecturestartinginthe1970’s.Thefollowingarethevariousformsofarchitecturefor thelastfiftyyears: • Mainframe(1970s) • PersonalComputers(mid-1980s) • Client-Server(late80stoearly90s) • WideWebandWeb2.0(mid-90stopresent) • CloudorPost-PC(todayandbeyond) Todaythereareavarietyofsoftwaresystemsavailabletobusinesses.Intermsofarchitectural format,theycanchoosebetween“OnLocation”or“OnPremise”–meaningthatthefirmshost thesoftwareintheirownfacilitiesorontheirhardwareandwithintheirownfirewall.However, companiesareincreasinglyoptingforcloudcomputingoptions.Thismeansthattheyhave severaldeploymentmodelsavailabletothem. Cloud Computing Ascloudcomputingbecomesincreasinglymorepopular,thereareavarietyofofferingsthat canbetailoredtofirmneeds.TheyareInfrastructureasaService(IaaS),PlatformasaService (Paas),andSoftwareasaService(Saas).Wediscusseachformataswellasbenefitsbelow: InfrastructureasaService(IaaS):Inthisformat,thethirdpartyprovidesthefirmwiththe computinginfrastructure,physicalorvirtualmachinesandotherresources.Firmownsand managesthesoftwareapplication.Thebenefitsofthisare: • Noneedtoinvestinyourownhardware V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 42 • • Infrastructurescalesondemandtosupportdynamicworkloads Flexible,innovativeservicesavailableondemand PlatformasaService(PaaS):Inthisformat,thethirdpartyprovidesfirmcomputingplatforms toincludeoperatingsystem,database,webserveretc.Firmownsandmanagesthesoftware application.Thebenefitsofthisare: • Developapplicationsandgettomarketfaster • Deploynewwebapplicationstothecloudinminutes • Reducecomplexitywithmiddlewareasaservice SoftwareasaService(SaaS):Inthiscase,thethirdpartyprovidesfirmwithaccesstothe applicationsoftwareandhandlesinstallation,setup,maintenance,andrunning.Firmis chargedbyuse.Benefitsinclude: • Youcansignupandrapidlystartusinginnovativebusinessapps • Appsanddataareaccessiblefromanyconnectedcomputer • Nodataislostifyourcomputerbreaks,asdataisinthecloud • Theserviceisabletodynamicallyscaletousageneeds Whiletherearemanybenefitstocloudcomputing,therearewidespreaddisadvantagesof cloudcomputingthatincludebutarenotlimitedto:vendoroutages,unrestrictedgovernment access,security&privacyrisks,andkeydataandprocessesrequirenetworkaccess. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 43 Software Selection Sources Therearedifferentsourcesofsoftwarethatfirmsneedtobeawareof.Theseincludea customizedin-housesystemdesignedforabusiness,anERPexpandedsystemwithadditional bellsandwhistlestailoredforacompany,bestofbreedsolutions(ofmarketsolutions),and bestofbreedplatforms.Thesearediscussedinfurtherdetailinthechartbelow: ProsandConsforSo3wareSources Source Customized In-House System ERP Expanded Systems Advantages Disadvantages • Bestfittothefirmanditsprocesses. • Excep7onallydifficultand7me consumingtodevelop • Mostexpensivetotalcostofownership • Difficulttomaintain • Canresultin“inwardlooking”solu7on • Rela7velyfastimplementa7on • Lessexpensivethanin-house customiza7on • EfficientfromITperspec7ve • EasiertoupgradewithERP enhancements • Tendstobeinflexibleintermsof process • Couldrequirechangeinbusiness processes • Notguaranteedtobebestsolu7on approach BestofBreed • Bestperformingmarketsolu7onforeach • Difficulttointegratedifferentsystems Solu7ons func7on • Canhaveslowperformance • Requirestheuseofmiddleware betweentheapplica7ons • Upgradingindividualcomponentscan causerippleeffectproblems BestofBreed • Verygood,ifnotbest,solu7onforeach • Requirestheuseofmiddleware Plaborms func7onwitheasierintegra7onbetween betweentheapplica7ons individualmodules MIT Center for Transportation & Logistics Outsourcing 14 Thereisalsotheoptionofoutsourcingsomeofthesesystemstodifferentproviders.For instance,inlogistics,3PLsorThirdPartyLogisticsProviders,serveasanorganizationthatcan runthesoftwareaswellasperformallofthebusinessprocesses.Havinga3PLrunyourlogistics eliminatestheneedforhardwareandsoftware.3PLscanpossiblyreplacepersonnelwithinthe firm.Theuseof3PLsismostcommonwithsmallerfirms. Themainreasonstooutsourcearetoreducecapitalexpenditureforsoftwareandhardware.It mayalsoreducecostsasaresultofpartner’seconomiesofscale;theyoftenhavetheabilityto doitfasterandbetteraswellasbemoreflexibileandagile.Itmayalsobeanopportunityto increaselevelsofserviceatreasonablecosts.Firmcanfocusoncorebusinessandbringin expertisethatisnotaffordablein-house.Therearemyriadotherreasonstooutsource,but therearealsomanynotto,discussedbelow. Atthetopofthereasonsnottooutsourcearesecurityandprivacyconcerns;someoneelsehas accesstothefirm’sdata.Therearealsoworriesofvendordependencyandlock-in.Thefirm maylosein-houseexpertisetoacorefunction.Therearealsohigh-migrationcostsaswellas concernsoveravailability,performance,andreliability.Thereareadditionalreasonsnotto V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 44 outsource.Firmsneedtoweightheprosandconsofoutsourcingmatchedwithbusiness objectivestodecidewhichissuitableforthem. Software Vendor Selection Intheend,afirmmustselectitsvendor.Somefirmsthrowadartonthewall,andthatisits choice.Othershaveanorganizedandformalizedfashiontoselectavendorandingeneralit goesasfollows: 1. FormaProjectTeam(Internaland/orExternal)&Objectives 2. UnderstandtheBusinessandNeeds:reviewcurrentbusinessprocesses,prioritize needs/functionality,createRequestforInformation(RFI) 3. CreateInitialShortListofPotentialSolutions&Vendors 4. In-depthReviewofShortListedVendors:havevendorsconductrealisticproduct demonstrations,referencesfromcurrentusers. 5. CreateandDistributefinalRequestforProposal(RFP) 6. MaketheDecision:negotiatecontract,price,andservicelevelagreements(SLAs)and establishanimplementationplanplan Whilecostisoneoftheprimaryfactorsindecision-making,therearemanyothercriteriathat needtobeevaluatedontopofcost.Theyare: • Functionality–doesthesystemfeaturesfitthefirm’sprocessesandneeds? • EaseofUse–howfastistheinitiallearningcurveandon-goinguse? • Performance–whataretheprocessingspeeds? • Scalability–howwellcanthesystemexpandandgrowwiththefirm? • Interoperability–howwelldoesthesystemintegratewithothersystems? • Extendibility–howeasilycanthesystembeextendedorcustomized? • Stability–howreliableisthesystemintermsofbugsandup-time? • Security–howwelldoesthesystemrestrictaccess,controlconfidentialdata,and preventcyberhacking? • Support–howisthequalityofthevendorintermsofimplementation,support, training,thoughtleadershipetc.? • VendorViability–howisthevendor’sfinancialstrengthandwillingnesstosupply updatesandenhancements?Willtheybeherein3years?? Becausethereareavarietyofcriteriathatfirmswillbeevaluatingvendorson,ascorecardisa popularwaytocapturefinancialandnon-financialattributes.Thecriteriacanbescoredasrank, ratings,andgrades.Scorecardstendtobeverydetailedandcanevenbebrokendownby specificfeatures.Theselectioncanbemadebetweenvendorsorbetweenalternativehosting platforms. Total Cost of Ownership SoftwareLicense:Directcostofthesoftwaresystemitself–assumingownership. Maintenance:Ongoingannualcoststoguaranteeupgradesandbugfixes. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 45 Platform/Hardware:Costofneededhardwaretorunthenewsoftware. Training:Costoftraininginitialandongoingpersonnel Implementation:Costofgettingthesystemtogolive!Thesevarywidelybetweensystemsand firms. Customization:Costofmodifyingthesystemitselftofitthefirm’sprocesses.NothinginSCMis usedstraightoutofthebox(vanilla). SystemIntegration Costofinterfacingthissystemwithothermodulesandmodifyingexistingsystemstofit Implementation Whileselectingavendorcanbedifficultandtimeconsuming,theactualprocessof implementationcantakeanevenmoresignificantamountoftimeandconsumealotof resources.Thereareafewdifferentapproachestoimplementation.TheyincludeDirect(orBig Bang),Parallel,Pilot,andPhased(orRolling).Eachofthesehasitsownpositivesandnegatives, buttheapproachmustsuittheneedsofthebusiness. Movingfromanoldmul7-modulesystemtoanewmul7-modulesystem. Implementa7onApproaches #Modules #Loca7ons Comments Converted Converted Direct or BigBang All All • Switchfromtheoldtonewsystemoccursononeday • Painofswitchconcentratedforen7refirm • Fastestimplementa7on7me,buthighestrisk • Post-implementa7onproduc7vitydrop • Highpoten7alforsystemwidefailuresduetoinsufficienttes7ng/training Parallel All/Some All/Some • Oldandnewsystemskeptonfortes7ngperiod • Lowestriskoffailure,buthighestcostandlongestimplementa7on7me • Employeesdodoubleentrywork Pilot All One Phased or Rolling One All • Fullimplementa7onofallmodulesatoneloca7on • Iden7fybugsorissuesthatarecorrectedpriortolargerrollout • Containsanypoten7alfailurefrominfec7ngallloca7ons • Testsindividualmodulesandintegra7onsimultaneously • Implementa7onofonemoduleata7meacrossthenetwork • Longerimplementa7ondura7onthandirect,butwithlowerrisk • Usershavemore7me&learnastheygo-nodipinperformancea3er • Learnandfixasyougo–beserprocessforlaterimplementa7ons • Lossofmanagerialfocusover7meandacon7nuousstateofchange • Poten7alformissingdataduringtransi7onalimplementa7onperiod • Mightrequiretemporarybridgesfromoldtonewsystemsduringtransi7on MIT Center for Transportation & Logistics 27 Thereareafewbestpracticestokeepinmindwhengoingaboutimplementation.They include: 1. Secureseniorexecutivecommitment:abilitytogatheranduseresources,empower team. 2. Forminterdisciplinaryteam(s) 3. Createaclearandspecificscopedocument 4. Buildextensivetestingintotheprojectplan(youcan’ttesttoomuch) 5. Includeextensiveusertrainingintotheprojectplan V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 46 Learning Objectives • • • • • • Recognizeselectingasoftwarevendorisanintertwineddecisionbetweenarchitecture andsource. UnderstandtradeoffsbetweenOn-PremiseandCloudbasedsystems. KnowthedifferencesbetweenIn-House,BestofBreed,ERPExtensions,andOutsourced formsofsoftwaresystems. Reviewtheselectionprocess,recognizingtherearemultipleattributes,andthetotal costofownershipiscomplex. Understandthechallengeofimplementationandthevariousapproachesto implementingsystemswithinafirm. Reviewbestpracticesofimplementation. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 47 Technology Trends Summary Asfuture(orcurrent)supplychainprofessionals,youwillbeconstantlyfacedwithtechnology trendsthatwillinfluenceyourworkandfutureplanning.Manyofthetoolswehavearmedyou witharetriedandtruemethodsandmaynotchange.Others,suchasvarioussystemsand technologieswillchangecontinuously.Itisimportanttobeawareofafewemerging technologytrendsweseetoday(asof2016).Wewillintroduceyoutoafewoftheseinthe nextlesson.Eachsectionwillcoverwhatthetrendis,itsstatus,andpotentialimpact. Trend: Autonomous Vehicles Autonomoustrucksaretrucksandothervehiclesthatcanoperatewithminimal(orno)human interaction.TherearevariouslevelsofAutomationfromNone(0)allthewayuptoLimitedSelfDrivingAutomation(Level3).Thestatusasoftodayisthatautonomousdeliveryhasalready happened.ThefirstpaidautonomousdeliveryoccurredinColoradoinOctoberof2016;Otto deliveredafullTLofbeer.TherehavebeensignificantinvestmentsintechnologysuchasUber acquiringOttoin2016for$680M.IntelacquiredMobileyein2017for$15B.Thedirectimpact ishuge:asingledayrangeoftruckscoulddoubletoabout1000miles.Itcouldlowerfuelcosts duetolowerspeeds. Approach § Networkmodelingwithnewtransportcharacteristics § Truckingindustrystrategy(Porters5ForcesModel) Trend: Delivery Drones Youhavealmostdefinitelyheardofdeliverydronesasthefutureofdelivery.DeliveryDrones areanunmannedaircraftthatcannavigateautonomously,withoutdirecthumancontroloris guidedremotely.Itwasusedinmilitaryoperationsstartingin2000.DroneDeliverieshave alreadyhappenedsuchasonDec.72016,AmazonPrimedeliveredanAmazonFireTVanda bagofpopcornbydronetoamannearCambridge,UK.Flirteyand7-Elevendeliveredachicken sandwich,donuts,candy,SlurpeesandhotcoffeeviadroneinJuly2016inRenoNV. Directimpactsincludetheabilitytosendsmallloadstoremotelocationsquickly.Itexpands deliverycapabilitiesusingopenairwithoutusingexistinginfrastructure.Itwillessentiallyserve asanewtransportationmodeforveryfastreplenishmentofverysmallshipmentsizeovera closedistance. AnalysisApproach § Networkmodelingwithnewtransportcharacteristics V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 48 § Demandplanning–newserviceoffering Trend: Mobile Computing Thenextmajortechnologytrendismobilecomputing.Itisatechnologythatallows transmissionofdata,voiceandvideoviaacomputeroranyotherwirelessenableddevice withouthavingtobeconnectedtoafixedphysicallink.Todaymobilecomputinghasgrown18 foldoverthepastfiveyearsofuse.Thereare8billionmobiledevices(including325million wearabledevices)currentlyinuse.Smartphoneusagegrewin38%in2016,buttheyareonly usedasphones3%ofthetime. Immediatedirectimpactsareobvious.Weuseitconstantly.Anyonewithasmartphoneisable toaccessdataandsystemsfromanywhereatanytime.Forsupplychains,itmeansnew paradigmsofshippingandretailsupplychainssuchasOmnichannel. AnalysisApproach § Demandplanningmodels–impactoflocationandsignalsforbetterforecasting § Inventorymodeling–changingstockingpointsandfunctions § Transportationmodeling–deliveryformorediverselocations § Warehousing–changingthefunctionoffacilities Additive Manufacturing Thefinaltrendwewilltalkaboutinthislessonisadditivemanufacturing.Thereareofcourse manyothersthatyouwillcrossinyourpathasasupplychainprofessional;thesearejustafew toshowwhatiscomingdownthelinecurrently.Additivemanufacturingistheprocessof makingaphysicalobjectfromathree-dimensionaldigitalmodel,typicallybylayingdownmany thinlayersofamaterialinsuccession.Asof2016additivemanufacturinggrew26%to$5.16B. Therearemanydifferentprocessesthatincludebutarenotlimitedto:materialextrusion, materialjetting,binderjetting,sheetlamination.Materialsincludepolymers,composites, metals,ceramics,paper,andmore.Ithasexpandedbeyondhobbyandprototypinguses.A primaryexampleisGEacquiringSwedenbasedArcamandGermanybasedSLMsolutionsfor $1.4Bin2016. Learning Objectives • • • • Technologyisalwayschanging! Assupplychainprofessionalsyouwillbecontinuouslyfacedwithnewtechnologytrends anddecidingwhatisviable. Beintroducedtocriticalthinkingabouttechnologytrends,theirpotentialdirectand indirectimpacts,andtimelines. Reviewcurrenttechnologytrends. V1 Spring 2017・CTL.SC4x – Technology and Systems・MITx MicroMasters in Supply Chain Management MIT Center for Transportation & Logistics・Cambridge, MA 02142 USA ・[email protected] This work is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License. 49