Forecasting Dropout of Undergraduates Pibulsongkram Rajabhat University with Data Mining Technique

Abstract

The objectives of this research were 1) to analyze characteristics and study factors related to the dropout of undergraduate students. Pibulsongkram Rajabhat University 2) to create a model and compare the efficiency of the model and 3) to develop a forecasting system for student dropout. The data of undergraduate students were obtained from the Education Services Division, Pibulsongkram Rajabhat University, between the 2015 and 2020 academic years, a total of 20,093 data sets containing 16 attributes. The filter approach feature selection method based on information gain was presented to analyze the factors affecting student dropout. It was found that ten factors influence student dropout, namely total GPA, year attended in university, programs of study, course of study, mother's occupation, father's occupation, faculty, educational level, region, and aptitude/talent. Then, the relevant factors were used to create the model using 3 data mining techniques: Decision Tree, Naïve Bayes, and Rule Induction technique. The model's performance was investigated using the 5-Fold Cross-Validation and 10-Fold Cross-Validation methods. The model's accuracy and mean absolute error (MAE) were also quantified. The results showed that the decision tree technique model had the highest value in the 10-Fold Cross-Validation. The accuracy and MAE of the model were 97.81% and 0.026, respectively. The researcher also designed and developed the model system using the form of a web application. The results from system evaluation data found the total mean of the correct prediction of the system. Accounted for 86.29%.