Predicting Student Performance: Analyzing Socio-economic and Personal Factors

Abstract

The purpose of this study is to predict student performance in Mathematics, Reading, and Writing exams by analyzing a comprehensive dataset containing student profiles and exam scores. Python's data processing and machine learning libraries are utilized for data analysis and modeling. The dataset is explored, followed by data preprocessing that includes feature encoding and standardization. Support Vector Regression (SVR) is employed as a multi-target regression model to train and evaluate the model for predicting student exam scores. Feature importance analysis and visualization of prediction errors are performed to gain deeper insights into student performance. The results demonstrate significant correlations between factors such as gender, lunch type, ethnic group, and parent education level with exam scores. The prediction model performs well on the test dataset and effectively predicts student exam scores in different subjects. The findings of this study provide valuable insights for educational decision-making and promote personalized and effective education.