Efficiency of Hybrid Decision Tree Algorithms in Evaluating the Academic Performance of Students

Abstract

Educational institutions are anticipated to take substantial and proactive roles in guaranteeing students' successful program completion. Academic performance is conventionally employed to categorize and forecast students' future ability to confront post-graduation challenges. A student's academic accomplishments are instrumental in shaping exceptional individuals who may become future leaders. Using algorithms to assess and predict academic performance is a wellestablished practice in machine learning, encompassing techniques such as neural networks , logistic regression , decision trees , and others. The goal of this project is to improve decision trees' ability to predict students' academic achievement via the use of data mining methods and metaheuristic algorithms. Educational data mining involves the utilization of data analysis methodologies and tools to examine the extensive data generated within educational establishments as a result of students' interactions and activities throughout their academic journey. Pelican Optimization Algorithm (POA) and Runge Kutta optimization (RKO) are utilized algorithms in developing hybrid models, both of which can efficiently search for optimal or near-optimal splits by fine-tuning the hyperparameters of decision tree models. Students' final grades were predicted through training and testing models and categorized into four classes: Excellent, Good, Acceptable, and Poor. The classification capability of a single model and optimized counterparts was evaluated using Accuracy, Recall, Precision, and F1-score in separate phases for each category. Obtained results for all models revealed that POA and RKO developed Accuracy of DTC by 1.86% and 0.87%. Also, Precision and Recall metric analysis further manifest the superiority of DTPO. Prediction based on classifiers, especially workable optimized versions such as DTPO, paves the way for institutions to raise student success rates.