Predicting Students Dropout using ensemble machine learning methods and hyperparameter optimization

Abstract

Education is the backbone of any economy. For any developing country especially India, education is very important. Education makes a person stable and prosperous in his way of leading life. In the same way, the number of higher educated persons in a country can contribute to the development and progress of the country. However, this number decreases due to the dropout of learners without completing the academic course. The dropout rate refers to the percentage of students who leave an educational program before its completion. Various reasons contribute to the dropout percentage, such as academic difficulties, personal or family reasons, financial problems, and lack of interest in studies. Dropout rates vary by region, socio-economic status, gender, and level of education in India, also students from lower socio-economic backgrounds are more likely to drop out due to financial constraints or the need to work to support their families. A student completing the course acquires improved skills and knowledge in comparison to the students who are leaving the course before completion. A machine learning-based predictive model is developed to classify students who are opting out the university course before its completion. A students’ dataset is considered to build the predictive model that consists of essential information about the students. Data pre-processing followed by statistical-based feature selection is performed on the dataset. The machine learning algorithms such as Support Vector Machine/SVM, K-Nearest Neighbour/KNN, Random Forest/RF, Decision Tree/DT, XGBoost, Naïve Bayes/NB, and Logistic Regression/LR are applied to build the classifiers. After building the classifiers, hyperparameters are adjusted to achieve classification accuracy.