AN IN-DEPTH STUDY ON STUDENTS’ PERFORMANCE EVALUATION USING MULTIPLE MACHINE LEARNING CLASSIFIERS AND DATA ANALYTICS APPROACHES

Abstract

Over the years, institutes have shown interest in improving the quality of education. To enhance education quality, it is important to predict whether a student is at risk of dropping out and to identify if they will achieve high or low scores. High dropout rates and low academic performance not only affect individual students' futures but also impact the institute's reputation. Improving academic performance and student retention involves collecting relevant data to identify students at risk. We use a Kaggle dataset in our research for prediction purposes. The data collected from students includes background information, academic records, behavioral data, and institutional details. Different preprocessing methods are used to improve data quality and validate models that predict student dropout and performance at an early stage. Various machine learning algorithms, such as Decision Tree, Random Forest, Logistic Regression, and Support Vector Machines, are used to evaluate models and analyze student datasets that produce overall results in academic performance. We evaluate all models on both training and testing data and also calculate their accuracy, comparing all model results for better prediction. By applying the Hist Gradient Boosting Classifier, accuracy reaches 90.7%. It is an updated version of the Gradient Boosting algorithm that enhances the efficiency of Gradient Boosting, making it suitable for large datasets. The purpose of the label encoder is to encode categorical features into numerical values. Using the label encoder, accuracy reaches 96.8%, which shows the best performance.