Data Driven Prediction of Student Performance: Integrating Academic, Social, Personal and Behavioral Factors in Pakistan Higher Education

Abstract

Student performance is an important aspect of any educational institution. Several factors affect a student’s performance, e.g., learning assessments, co-curricular activities, previous achievements, personal and social Interactions, etc. Currently, in Pakistan, there is a lack of any such system that helps to analyze and predict student performance. There are two primary reasons for this, 1) lack of research into existing prediction methods to determine the best suitable approach for evaluating student performance in Pakistani institutions, 2) limited research on the factors that influence specifically Pakistani students' performance. The essential goal of this research is to foster predictive models for GPA evaluation using machine learning (ML) and deep learning (DL) techniques. The main research problems explore the efficiency of various features and algorithms in predicting students' cumulative GPAs (CGPA) and pass/fail outcomes. To close this gap, we first gather a sizable dataset of several characteristics that influence a student's academic success in this study project. The COMSATS University, Lahore Campus students studying computer science and software engineering are the source of the data. Additionally, we assessed the suggested dataset using (ML & DL) models to forecast students' performance in terms of regression (CGPA) and classification (pass/fail). The assessment is conducted utilizing f-score, accuracy, precision, and recall. The examination of RMSE, R2 , and MAE is used to assess the performance of regression techniques. The results of this study might be advantageous to academic institutions, teachers, and students. The study concludes by promoting the use of predictive models to backing academic interferences and institutional planning, with the potential for wider approval in several educational situations