Transcript Diapositiva 1

Improving quality of educational processes providing
new knowledge using data mining techniques
Manolis Chalaris, Stefanos Gritzalis, Manolis Maragoudakis, Cleo Sgouropoulou and Anastasios Tsolakidis
Technological Educational Institute of Athens, Ag. Spyridonos, 12210 Aigaleo, Athens, Greece
University of the Aegean, Department of Information and Communication Systems, Samos GR-83200, Greece
Presentation Outline
 Introduction
 Data Mining in Higher Education Institutes
 Case Study - Application of Data Mining Techniques in TEI of Athens
 Conclusion - Future Work
Introduction
Our aim in this work is
1. to demonstrate the ability of data mining in improving the quality of
educational processes and thus supporting the administration of an
educational institute in the decision-making procedure
2. to conduct some experiments in using data mining techniques like clustering
analysis, correlation analysis and association rules in educational data of the
Technological Educational Institute of Athens (TEIA)
Educational Data Mining: describes a new research field concerned with
developing methods for exploring the unique types of data that come from
educational settings, and using those methods to better understand students,
and the settings which they learn in.
Related work
– Romero and Ventura, survey between 1995 and 2005 to present application of
data mining techniques in different types of educational systems and then grouping
them by task.
– Al-Radaideh et al. applied decision tree as classification method to evaluate
student data in order to find which attributes affect the performance of a courses.
– Mohammed M. Abu Tair & Alaa M. El-Halees use educational data mining to
discover knowledge that may affect the students’ performance
– Baradwaj and Pal applied decision tree for evaluating students’ performance
– Karel Dejaeger et al. investigated the construction of data mining models to identify
the main attributes of students satisfaction and
– Dursun Delen examines the institution-specific nature of the attrition problem
through models developed by educational data using machine learning techniques.
Benefits and Success Factor of EDM
Benefits
• Extracting knowledge that will support the HEI administration in decision making
procedure for improving the quality of educational processes.
• Clustering can offer comprehensive characteristics of students, while Prediction
(classification and regression) and Relationship Mining (association, correlation,
sequential mining) can help the university to decrease student's drop-out rate or to
increase student's retention rate and learning outcome.
• Provide more personalized education, maximize educational system efficiency,
and reduce the cost of education processes
Success Factors
 Existence of an appropriate infrastructure for finding and collecting all educational
data in a centralized system (Q.A.I.S. of MODIP TEIA)
 Analysis Model as a roadmap for the institute to identify which part of the
educational processes can be improved through data mining and how to obtain
each strategic goal
Methodology: CRISP-DM methodology- the CRoss Industry Standard Process for Data
Mining.
The methodology consists of six steps or phases:
1. Business (Organizational) Understanding. This phase focuses on understanding the
project objectives and requirements from a business or organizational perspective.
2. Data Understanding where initial data is collected, data quality problems are identified
and/or interesting subsets to form hypotheses for hidden information are detected.
3. Data Preparation. In this phase all necessary tasks like data cleansing, data
transformation and data selection are performed in order to construct the final dataset.
4. Modeling phase where various modeling techniques are selected and applied
5. Evaluation phase in which you determine how valuable your model is, and if it
achieves the business objectives you have set.
6. Deployment phase. In this phase we have the actions that should be carried out in
order to use the created models.
Ultimate target concerning TEIA
Quality improvement of Educational Processes
In the framework of this paper,
the main objective is to conduct some experiments in data of the
evaluation process of the institute using data mining techniques and
thus extract knowledge concerning aspects of educational
processes.
In this research
– we focus on data collected by QAU of TEIA (MODIP TEIA) for the evaluation
process of the spring semester of the acad. year 2011 - 2012 between 8th –
10th week of courses.
– 2 questionnaires – theoretical courses with 35 questions (attributes) and
laboratory courses with 22 questions (attributes)
– The sections – directions of the 2 questionnaires are:
•
•
•
•
course-centred items (e.g. is the course well structured?)
lecturer & teaching effectiveness (e.g how s/he explains the content of the course)
student-centred items (e.g. how often you attend the course?) and
laboratory work (e.g. are the facilities of the labs sufficient?)
– sample of 26.000 questionnaires, approximately 10.000 for the theoretical
courses and 16.000 for the laboratory from students of 5 Faculties and 27 (33)
Departments
– Data Integration (Collecting all data in one dataset)
– Data Cleansing (Handling of missing values or inaccurate records)
– Selecting attributes for conducting the experiments
Cluster Analysis (k-means) to the data of
the theoretical questionnaire to examine
which faculty has better averages in
attributes of all three directions and
compare it with the percentage of graduates
of each Faculty
Faculty
Percentage of graduates –
Length of study between 4
and 5 years
Faculty of Management
and
Economics
52,86
Faculty of Technological
Applications
29.5
Faculty of Fine Arts and
Design
39.62
Faculty of Health and
Caring Professions
68,3
Faculty of Food
Technology and Nutrition
29.15
Correlation Analysis to examine if the student concepts' understanding
(q28_student_understanding in theory questionnaire) interacts with other
attributes
Association Rules (FP-Growth) to examine the student understanding (q17)
considering the laboratory courses.
Cluster Analysis (k-means) to the data of the laboratory questionnaire to
examine how students in the three Faculties (SEYP, SDO, STEF) evaluate
the facilities of the labs (q13 and q14).
Student Understanding relates mainly with the lecturer and teaching effectiveness
especially in the theoretical courses while in the laboratory courses, lab facilities
are considered as a premise for the student to understand. Students of the Faculty
of SEYP are more consistent with their studies (attendance, studying, and
understanding) than those of other Faculties. On the other hand, SEYP has the
biggest problem concerning the lab facilities.
•
•
Focus in the improvement of the lab facilities (input indicators) in SEYP (e.g
better resource allocation)
In other Faculties the focus should be mainly in the educational process
(process indicators) such as better organization of the course, improving
teaching effectiveness and student understanding (e.g Organizing pedagogical
seminars for the lectures in order to enhance their teaching effectiveness)
Present a more integrate approach of applying data mining techniques in a HEI.
1.Definition and Modelling all educational processes
2.Defining all indicators that determine the quality of them.
3.Proposal of an Analysis Model as guideline for the application of data mining in
educational data of TEIA
Create a strategic tool to support the decision making procedure for
enhancing the quality of educational processes
Thank You!
Manolis Chalaris
Quality Assurance Unit (MODIP) TEI of Athens
Email: [email protected]