An Efficient Approach for Multiclass Student Performance Prediction based upon Machine Learning

The field of educational data mining has enabled the researchers, educators to predict the student's pass rate, failure rate, dropout rate etc. The main reason for dropouts of the student is the failure of students. Several researchers have proposed various educational data mining techniques for predicting the student performance and analyzed existing techniques on educational datasets. In this paper, we have analyzed the performance of four machine learning algorithms on educational dataset used for the early prediction of student performance. While there is a rich literature survey in student performance prediction, our work differs from existing works as follows:(i) Our prediction is not limited to binary classification of pass and fail but we have used a multiclass classification in which student are divided into three classes namely poor, average, good performing students;(ii) We have used data preprocessing, feature extraction, fine tuning of parameters of algorithms to build the model which is not focused by many researchers;(iii) We have built a more accurate model with 95% accuracy and lesser execution time;(iv) We have studied the relationship between various attributes through a correlation heatmap. This paper will elaborate on student performance prediction techniques and give a comparative study among them.