An insight into Educational Data Mining at Muğla Sıtkı Koçman University, Turkey

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Educational Data MiningAn insight into EDM at Muğla Sıtkı Koçman University

Presentation by Steven Strehl, HTW BerlinE-Mail: [email protected]

June 30th, 2014

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Gürüler, Hüseyin and Istanbullu, Ayhan (2014): Modeling Student Performance in Higher Education Using Data Mining. In: Educational Data Mining, Alejandro Peña-Ayala (Ed.).

Gürüler, Hüseyin (2005): Veritabanları Üzerinde Veri Madenciliği Uygulaması. Muğla Sıtkı Koçman Üniversitesi.

Teşekkür ederim to Hüseyin Gürüler for providing support and additional information by mail and Sema Karakurt for helping me translate important aspects of Hüseyin’s Master’s thesis from Turkish to English.

Primary sources

Vision and Goals

Student Knowledge Discovery Software

Improve efficiency, quality, “experience” of studies

Predict failure or success of students

Eliminate factors that lead to failure

Figure 1: SKDS user interface

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Context

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Student profiling

Figure 2: Female student

Gender

Military service

Grade Point Average

Name

Secondary school

Scholarship

Age

Final mark

Secondary school

Nationality

Study programme

Family income

Marital status

Focus subject

Parents’ professions

Religion

Hometown

…

Context

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Figure 3: Individual studentsFigure 4: Anonymous students

Basics

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Knowledge Discovery in Databases

Data MiningAvailable Data Selected Data Pre-processed Data

Transformed Data

Interpretation and Evaluation

Figure 5: Stages of the KDD process

Basics

Data MiningCross Industry Standard Process

for Data Mining (CRISP-DM)

Verification-driven DMAims at verifying assumptionsby data queries

Discovery-driven DMAims at gaining new insightsby unveiling patterns

Figure 6: CRISP-DM Process Diagram

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Approach

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Data curation

Figure 7: Male student

Gender

Military service

Grade Point Average

Name

Secondary school

Scholarship

Age

Final mark

Secondary school

Nationality

Study programme

Family income

Marital status

Focus subject

Parents’ professions

Religion

Hometown

…

Approach

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Data volume and provenience

Data Mining

Microsoft Decision

Tree Algorithm

Available Data

University departments, faculties, central

registration system and archives

Selected Data

13 tables related to the scope of SKDS

Pre-processed Data

6 tables

Transformed Data

View consisting of 111 columns and

6,470 records

Interpretation and Evaluation

Figure 8: Stages of the KDD process

Approach

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Models and results

Model 1: GPA >= 3.0- Language prep. (English)- Registration preference

Model 2: GPA >= 2.0- Family income

Figure 9: Decision Tree for Model 1 Figure 10: Decision Tree for Model 2

Issues and Outlook

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Issues

Data availability, variety, format

Usability not yet suitable for everyday use

Transfer impact of findings

Outlook

Data Mining new algorithms available

MÜKÜP/SKDS was an early WEKA

Improvement new GUI, easier to use

Discussion

Except where otherwise noted, content of this work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Topics

Social background Family income has been discovered as animportant factor for success. Does the data scientist’s work end here?

Study conditions How to measure and find aspects that counterbalance negative preconditions of the social background?

Health How does the mental and physical state influence success and failure?

Open Data What if Student Life Cycle Data were open?

List of Figures

[1] Gürüler, Hüseyin and Istanbullu, Ayhan (2014): Modeling Student Performance in Higher Education Using Data Mining. No CC license.

[2] http://openclipart.org/user-detail/ryanlerch Public domain.

[3] http://openclipart.org/user-detail/ryanlerch Public domain.

[4] http://openclipart.org/user-detail/thekua Public domain.

[5] Composed graphic from http://openclipart.org/user-detail/jean_victor_balin, http://openclipart.org/user-detail/buggi, http://openclipart.org/user-detail/gsagri04 Public Domain.

[6] Kenneth Jensen, https://en.wikipedia.org/wiki/Cross_Industry_Standard_Process_for_Data_Mining#mediaviewer/File:CRISP-DM_Process_Diagram.png

[7] http://openclipart.org/user-detail/ryanlerch Public domain.

[8] Composed graphic from http://openclipart.org/user-detail/jean_victor_balin, http://openclipart.org/user-detail/buggi,http://openclipart.org/user-detail/gsagri04 Public Domain.

[9] Gürüler, Hüseyin and Istanbullu, Ayhan (2014): Modeling Student Performance in Higher Education Using Data Mining. No CC license.

[10] Gürüler, Hüseyin and Istanbullu, Ayhan (2014): Modeling Student Performance in Higher Education Using Data Mining. No CC license.