Download Intelligent Data Analysis and Data Mining Data Analysis and

Lluis Belanche + Alfredo Vellido
Intelligent Data Analysis and Data Mining
or …
Data Analysis and Knowledge Discovery
a.k.a. Data Mining II
An insider’s view …
Geoff Holmes: WEKA founder
Process Mining
IDADM
Process Mining (PM)
PM sits between CI and DM on the one hand, and process modeling and analysis on the other. PM aims to discover, monitor and improve real processes by extracting knowledge from event logs. event logs
Whyare extracted from data sources (e.g., databases, transaction logs, PM? … an ever‐increasing number of events are being audit trails, etc.). Examples of formats are
MXML (Mining eXtensible Markup recorded, providing detailed information about the history of Language)
and XES (eXtensible Event Stream). XES was selected by the IEEE processes. On the other hand, there is a need to improve and Task Force on Process Mining as the standard format for logging events.
support business processes in rapidly changing and aggressively There are several tools to extract MXML or XES logs from various data competitive environments.
sources. See for example:
PM includes (automated) process discovery (extracting process • XESame
• ProMimport
models from an event log), conformance checking (monitoring
• Nitro
deviations of model from log), organizational mining (inc. social networks), automated construction of simulation models, model extension, model repair, case prediction, and history‐based
recommendations.
IDADM
Process Mining (PM)
PM could be a bridge between DM and business process modeling and analysis, under the umbrella concept of Business Intelligence (BI). It can also be seen as the "missing link" between DM and traditional model‐
driven BPM. Most DM techniques are not fit as such for process analysis.
Co‐existing analytical concepts: Business Activity Monitoring (BAM): technologies enabling the real‐time monitoring of business processes. Complex Event Processing (CEP): technologies to process large amounts of Six Sigmaevents for optimizing the business in real time. Corporate Performance is a set of strategies, techniques, and tools for process improvement. It was developed Management
(CPM): measuring the performance of a process or by Motorola
in 1981.[and became famous when
it became a successful business strategy organization. at General Electric
in 1995. Today, it is used in many industrial sectors.
It seeks to improve the quality of process outputs by identifying and removing the causes of Co‐existing management concepts: such as Continuous Process
defects (errors)
and minimizing
variability in business processes.
quality Improvement
(CPI), Business Process
ImprovementIt uses a set of
(BPI), Total Quality management methods, including statistical methods
Management (TQM), and Six Sigma. PM enables all these within a single Each Six Sigma project carried out within an organization follows a defined sequence of steps
framework.
and has quantified value targets, for example: reduce process cycle time, reduce pollution, reduce costs, increase customer satisfaction, or increase profits.
IDADM
Process Mining (PM)
Event logs: All PM techniques assume that it is possible to sequentially record events such that each event refers to an activity (a well‐defined step in some process) and is related to a particular case (a process instance). EL may store additional information about events: resource (person or device) executing the activity, timestamp of the event, or data elements recorded together with the event.
IDADM
Process Mining (PM)
Discovery: The first element of PM is discovery. A discovery technique takes an event log and produces a model without using any a priori information.
Conformance: The second is conformance: an existing process model is compared with an event log of the same process. Conformance checking can be used to check if reality/process, as recorded in the EL, conforms to the model and vice versa. Conformance checking can be applied to procedural models, organizational models, declarative process models, etc.
Enhancement : Extending or improving an existing PM using information about the actual process recorded in some EL. This third type of PM aims at changing or extending the a priori model.
IDADM
Process Mining (PM): perspectives
•
•
•
Control‐flow perspective: focuses on the ordering of activities. The goal of mining this perspective is to find a good characterization of all possible paths. The result is typically expressed in terms of a Petri net or some other process notation (EPCs, BPMN, or UML activity diagrams). Organizational perspective: focuses on information about resources hidden in the event log, i.e., which actors (people, systems, roles, or departments) are involved and how are they related. The goal is to either structure the organization by classifying people in terms of roles and organizational units or to map a social network. Case perspective: focuses on properties of cases. A case can be characterized by its path in the process or by the actors working on it.
Business Process Model and Notation (BPMN) example. A graphical representation for specifying business processes in a business process model.
IDADM
Process Mining (PM): BPM vs. PM
•
Business Process Modeling: 7 phases : In the (re)design phase a new process model is created or an existing process model is adapted. In the analysis phase a candidate model and its alternatives are analyzed. Then, the model is implemented (implementation phase) or an existing system is (re)configured
(reconfiguration phase). In the execution phase, the designed model is enacted. During the execution phase the process is monitored. Moreover, smaller adjustments may be made without redesigning the process (adjustment phase). In the diagnosis phase the enacted process is analyzed and the output of this phase may trigger a new process redesign phase.
IDADM
Process Mining (PM): BPM vs. PM
PMining: 5 stages :
Plan and Justify: Includes understanding the available data and process domain. Extract: event data, models, objectives, and questions need to be extracted from systems, domain experts, and management. Control‐flow modelling: control‐flow model is
constructed and linked to the event log. Here
automated process discovery techniques can be used. The event log may be filtered or adapted using the model (e.g., removing outlier cases and inserting missing events). Integrated process model: the control‐flow model may be extended with other perspectives (e.g., data, time, and resources).
Operational support: Moreover, smaller adjustments may be made without redesigning
the process (adjustment phase). In the diagnosis phase the enacted process is analyzed and the output of this phase may trigger a new process redesign phase.
IDADM
Process Mining (PM): Guiding principles
PMining: 5 stages :
Plan and Justify: Includes understanding the available data and process domain. Extract: event data, models, objectives, and questions need to be extracted from systems, domain experts, and management. Control‐flow modelling: control‐flow model
is constructed and linked to the event log. Here automated process discovery
techniques can be used. The event log may be filtered or adapted using the model (e.g., removing outlier cases and inserting missing
events). Integrated process model: the control‐flow model may be extended with other perspectives (e.g., data, time, and resources).
Operational support: Moreover, smaller adjustments may be made without
redesigning the process (adjustment phase). In the diagnosis phase the enacted process is analyzed and the output of this phase may trigger a new process redesign phase.
IDADM
Process Mining (PM)
PM as a building block of BI
IDADM
Process Mining (PM)
PM book
IDADM
Process Mining (PM)
PM IEEE Task Force