DETECTION OF STUDENTS AT RISK OF ATTRITION USING DATA MINING APPROACH
DETECTION OF STUDENTS AT RISK OF ATTRITION USING DATA MINING APPROACH
No Thumbnail Available
Date
2016-08
Authors
ABDULSALAMI, AMINU ONIMISI
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Educational data mining (EDM) is a field that concentrates on prediction and is known for its role in uncovering hidden information from large volume of data. EDM has seen an emergence of research leading to strategies that aim to address issues of higher education with primary focus on students’ attrition. An unceasing student’s attrition from the undergraduate programme of Ahmadu Bello University, Zaria (A.B.U, Zaria) has significant ramification for individuals affected by this problem and the society in general. This work examines the problem of predicting student’s dropout in selected programmes of A.B.U Zaria. This work also focuses on addressing students’ attrition by creating a first year at risk model to be used in early detection of students at the risk of attrition. Various factors that might influence students’ attrition were collected from two (2) primary sources (student portal and exam processing software) of A.B.U Zaria information system. This work applies variants of existing classification algorithms in building predictive models; also an instance based learning algorithm (k nearest neighbor) was chosen and modified. An initial data preprocessing, feature selection and a 10-fold cross validation experiment was carried out in the model development phase of this work using WEKA (an open source tool for data mining tool). The classification algorithms employed in this work include: Multi-Layer Perceptron, Naïve Bayes, J48 Decision Tree, Sequential minimal optimization, K-Nearest Neighbor and a modified K-Nearest Neighbor. Results obtained showed J48 decision tree algorithm have performed nicely among all, based on our dataset with an average accuracy of 97.9%. Also the modified nearest neighbor algorithm performed next with an average accuracy of 97.3%. The result obtained were validated and analyzed using WEKA’s experimenter. To choose the best model, we conducted a comparative analysis of the classifiers used in this dissertation work.
Description
A THESIS SUBMITTED TO THE SCHOOL OF POSTGRADUATE STUDIES
AHMADU BELLO UNIVERSITY (ABU), ZARIA
IN PARTIAL FULLFILLMENT OF THE REQUIREMENTS FOR THE AWARD OF MASTER OF SCIENCE (M.Sc.) DEGREE IN COMPUTER SCIENCE
DEPARTMENT OF MATHEMATICS
AHMADU BELLO UNIVERSITY (ABU),
Keywords
DETECTION,, STUDENTS,, RISK,, ATTRITION,, DATA MINING APPROACH,