Student academic performance prediction via artificial intelligence using machine learning algorithms

The academic success of students in schools is valued by both students, teachers, and families. For this reason, performance prediction plays a significant role in students' life. With performance prediction, unsuccessful students can be directed to increase their success, study programs can be created, supportive course resources can be suggested, or elective courses can be selected. In this study, the academic success of the student can be predicted with machine learning methods. This study made use of dataset collected from student's knowledge from two schools in Portugal from Kaggle web site. We used three different algorithms to make performance prediction. These are Decision Tree, Random Forest and Logistic Regression. 30% of the dataset was used as test data. The remaining 70% data was used as training data. By using three algorithms, the confusion matrix, accuracy, recall, precision and auc values are obtained. It has been concluded that which algorithm is more successful on which amount of data. decision tree algorithm gives the best accuracy rate with max depth 2 value with 649 student data. The random forest algorithm gives the best accuracy with 649 student data. The logistic regression algorithm gives the best accuracy with 395 student data.