Machine Learning Approaches for Student Performance Prediction

Abstract

India's Education system is very old and due to a large population of students in India, there are some serious issues in analyzing and predicting students' performance. In the Indian Context, every institution has its own set of standards for evaluating student success, there is no proper procedure for monitoring and analyzing a student's performance and progress. One of the major factors is lack of research in existing prediction approaches, making it difficult to determine the optimal prediction methodology for visualizing student academic growth and performance. Another reason could be the lack of research into the areas that can affect students' academic performance and achievement. In this paper, focus is given on additional external factors like geographical location, parent education, health status etc. that can affect a students' performances apart from the grades in any course. That will be more effective in visualizing and analyzing student's performance. For experimental work, data has been collected from UCI repository and results are obtained from two different machine learning algorithms (KNN and Logistic Regression). Performance analysis is also done for these two algorithms based on accuracy level of results as well as with some existing work.