Early Detection of Failure Risks from Students' Data

• Published in: 2020 International Conference on Emerging Trends in Smart Technologies (ICETST) pp. 1 - 6
• Main Authors: Kouser, Faiza, Meghji, Areej Fatemah, Mahoto, Naeem Ahmed
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.03.2020
• Subjects: Accuracy; At-risk students; Bayes methods; classification; Classification algorithms; Data models; Employee welfare; Organizations; prediction; Prediction algorithms; Predictive models; Random forests; WEKA
• DOI: 10.1109/ICETST49965.2020.9080692

In academic organizations, identification of at-risk learners as quickly as possible is a serious challenge. The goal of an educational institution is providing a learning environment that maximizes students' performance and identifies at-risk students at the earliest possible time. Early detection of at-risk students will lead to a reduction in failure rates at the end of a semester. This research aims to provide a comparison of different classification methods such as lazy-based, ruled-based, tree-based, function-based and Bayes-based algorithms to predict at-risk students on the basis of descriptive data at an early stage of a semester. The students' found at risk for failures may be provided special counseling and tutoring before end of the semester. The study in hand, also compares classification algorithms utilizing their TP rate, FP rate, precision, recall, F-measure, accuracy, kappa statistics and model building time. The experiments have been performed using a real dataset, and results show competitive outcomes of several algorithms. The Random forest and Decision Table outperformed with 95.37% and 94.60% accuracy respectively.