College Dropout Prediction Model using Supervised Machine Learning

Authors

  • Diosdado C. Caronongan

Abstract

Analyzing corporate data to understand and find patterns in customer behavior is a critical exercise that should be undertaken by every business establishment in order to stay competitive. Accordingly, in the education sector, analyzing dropout data will help schools better understand student behavior. Education statistics show that foreign as well as local educational institutions are experiencing an alarmingly high dropout rates. Thus, a school equipped with a tool that can predict a dropout, will have a competitive advantage since it can timely prescribe needed programs of intervention that could prevent dropping out from happening.
This research work aimed to develop dropout prediction models using the two major types of supervised machine learning – Classification Method and Regression Method. The dataset used in the study are based on the academic records of 687 Information Technology and Computer Science students who are enrolled at University of Luzon, Dagupan City Philippines.
Using the combination of accuracy, precision, recall and F measure metrics, the study compared the prediction performances of the two supervised learning models and determined the better solution. Likewise, the effect of feature engineering on the performances of the prediction models were measured and determined. Moreover, a web-based dropout prediction system was developed and deployed using the Shiny package framework.

Downloads

Published

2020-03-27

Issue

Section

Articles