Machine Learning for Predicting Student Engagement and Adaptability in Online Courses

Abstract

The abstract summarizes the key findings and methodology of the study conducted on various machine learning classifiers for predictive analysis on a specific dataset. Through comprehensive exploration, classifiers like Decision Tree, Random Forest, and XGBoost emerged as high performers, achieving an average accuracy of around 90% with minimal misclassifications. Conversely, classifiers such as Logistic Regression, Gaussian Naive Bayes, and Multi-layer Perceptron demonstrated lower accuracies, indicating their limited suitability for the dataset. Additionally, hyperparameter tuning of the Decision Tree model using a grid search method led to a significant improvement in accuracy to approximately 90.87%. These findings underscore the critical role of thoughtful model selection, evaluation, and optimization in developing accurate and reliable machine learning models for real-world applications.