Download Proposal Presentation

Investigating mobile based prediction modelling of
academic performance for primary school pupils: a
data mining approach.
by Mvurya Mgala
Supervisors:
Dr Audrey Mbogho and Prof Gary Marsden
Introduction
Problem statement
• Low academic performance of primary school pupils
in some regions has presented a worrying trend.
• Research has shown this to be a widespread problem
among developing nations.
• The problem has been attributed to many factors,
ranging from students’ personal factors, teacher
factors, school factors, to family background factors.
Motivation
• The study will reveal the impact of the factors
surrounding pupils’ low academic performance.
• Discover the causes with the highest impact which
can be used to build a prediction model.
• The model will be used by education stakeholders to
predict pupil’s performance.
• Will propose intervention measures and facilitate
informed decision making.
Area of research
Population: 649,931
Divisions: Kinango, Matuga,
Msambweni, Kubo.
Area: 832,200 ha
What a contrast
Crowded classes
Lack of facilities
Poverty
So what is the impact of these factors
on academic performance in primary
schools in the developing world?
Research Questions
• How can the Bayesian classifier be modelled from
the primary schools’ data?
• How can a Bayesian model be used for prediction of
primary school pupils’ academic performance?
• What mobile application artefact can be designed to
automatically predict academic performance?
Significance of the study
Findings of the study will contribute to the field of
computer science and KDD in the following way:
• Provide a process to design and create a prediction
model artefact that predicts academic performance for
primary school pupils.
• Expose the social and technological issues that influence
the successful design, implementation and adoption of
an academic performance prediction model.
• Support and enrich the classification approaches in
implementation and adoption of prediction systems.
Research Approach
Gather data through semi-structured interviews,
questionnaires and secondary data,
Pre-process the data to extract relevant factors
that affect academic performance
Data mining: apply specific algorithms to extract
patterns from data,
Interpretation: making sense out of the extracted
patterns,
Knowledge: the sense made out of the patterns,
Data mining process
Interpretation
Data Mining
Transformation
Preprocessing
Selection
Patterns
Preprocessed
Original
Data
Target
Data
Data from semi-structured
interviews Questionnaires
and secondary data
Data
Transformed
Data
Knowledge will
be extracted by
stakeholders from
a mobile phone
Classification Algorithms
“What factors determine low academic
performance in primary schools?”
Data
Pupil, teacher,
School, parent data
Mining
Algorithm
General
patterns
Classification
Algorithm
Bayesian
networks
Patterns coded
into a mobile app
Methodology
Target Population
• The research will be targeted towards primary school pupils
of 10 primary schools in Kwale County in Kenya.
Sampling Design
• Stratified sampling will be used since the target group is
known.
• A list of the factors will be obtained through literature and
semi-structured interviews with 18 education officers.
• Questionnaires will be given to 50 teachers and 200 pupils.
Model evaluation-Confusion Matrix
Predicted
Original classes C’
B’
A’
Total
C
50
0
0
50
B
0
48
2
50
A
0
4
46
50
Total
50
52
48
150
• Rows show the actual classes
• Columns show the predicted classes
Mobile artefact evaluation
• The mobile application artefact will be
evaluated using a field-based usability
evaluation methodology
Impact evaluation
• Five pupils in their final year will randomly selected.
• The prediction artefact will be used to determine
their likely outcome in the final examination.
• Some intervention measures will be put in place and
the pupils’ final results compared with the predicted
grades.
• The study will therefore be able to propose possible
interventions to the stakeholders.
Contribution to Knowledge
•
Design and testing of a prediction model for academic
performance of primary school pupils.
• Provides an alternative to dependence on final
examinations to determine students’ abilities
•
Provide means by which decision makers can make
accurate decisions and effective policies.
The end!