Due to COVID-19, the researching of educational data and the improvement of related systems have become increasingly important in recent years. Educational institutions seek more information about their students to find ways to utilize their talents and address their weaknesses. With the emergence of e-learning, researchers and programmers aim to find ways to maintain students' attention and improve their chances of achieving a higher grade point average (GPA) to gain admission to their desired colleges. In this paper, we predict, test, and provide reasons for declining student performance using various machine learning algorithms, including support vector machine with different kernels, decision tree, random forest, and k-nearest neighbors algorithms. Additionally, we compare two databases, one with data related to online learning and another with data on relevant offline learning properties, to compare predicted weaknesses with metrics such as F1 score and accuracy. However, before applying the algorithms, the databases need normalization to meet the prediction format. Ultimately, we find that success in school is related to habits such as sleep, study time, and screen time. More details regarding the results are provided in this paper.
Keywords: decision tree; e-learning; k-nearest neighbors; machine learning; random forest; students’ performance; support vector machine.