Predicting Students Academic Performance using a Hybrid of Machine Learning Algorithms

Abstract

Educational data mining (EDM) has become a very interesting field of study in machine learning (ML), since it has enabled searchers to mine knowledge from educational databases for improvement in students’ and instructors’ performance. The most challenging task in prediction is to identify which features and algorithms to select which will give satisfactory results. In this research, a hybrid algorithm of weighted voting classifier (WVC) in conjunction with 10-fold cross validation (10-CV) and five other machine learning algorithms that are support vector machine (SVM), multi-layer perceptron (MLP), logistic regression (LR), k-nearest neighbor (KNN) and naive bayes (NB) were used. We evaluated our proposed model on the student grade prediction dataset taken from kaggle. In this paper, the metrics that were measured included: accuracy, precision, recall, f1-score and area under the curve (AUC). An accuracy of 97.6% was achieved. The proposed model was able to identify 634 students out of 650 as (Fair, Good, and Excellent), therefore recommending the model to the school for student performance prediction since it will devise mechanisms to alleviate student dropout rates and improve their performance.