A DATA-DRIVEN APPROACH TO EARLY STUDENT PERFORMANCE PREDICTION USING COLLABORATIVE FILTERING

Paper Title

A DATA-DRIVEN APPROACH TO EARLY STUDENT PERFORMANCE PREDICTION USING COLLABORATIVE FILTERING

Abstract

The education sector has emerged as a prominent area for data exploration, leading to an increased interest among companies in educational data. As a result, the demand for data-driven insights in this field has risen significantly. Through the application of data mining and machine learning techniques, I focused on extracting valuable information from educational datasets to develop automated tools aimed at enhancing the education domain. This initiative centers on analyzing student achievement by employing various machine learning and data mining methodologies. Real-world data, including student grades, demographic information, social factors, and school-related characteristics, was gathered from school reports and questionnaires. Four distinct data mining models were utilized: Decision Tree, Random Forest, Neural Networks, and Support Vector Machines. These models incorporated multiple variables that correlate strongly with final grades. The findings demonstrate a commendable accuracy when predicting outcomes based on students\' first, second, and third-grade scores. While past evaluations substantially influence students\' performance, a comprehensive analysis reveals that several additional factors, such as attendance rates, parents\' occupations and education levels, and alcohol consumption, also contribute to academic success. This research opens the door for the development of advanced student prediction tools and provides insights into the challenges faced by special needs students. Ultimately, these findings aim to improve educational quality and optimize school resource management.
KEYWORDS :