Impact of AI Based Learning on Student Performance Prediction in Higher Education Using Machine Learning

Prathamesh Deepak Salke, Dr. Ayesha Siddique

Abstract: Research on what influences student academic achievement has been ongoing for a while now; however, little has been documented about the impact of the use of artificial intelligence (AI). In this paper, we seek to bridge this gap by employing an empirical analysis of the effects of AI adoption using two machine learning methods, namely, Random Forest and Light Gradient Boosting Machine (LightGBM), on 8,000 student profiles. There were 22 variables representing students' performance, their learning habits, AI usage, and demographics. The two models are trained to classify students based on their performance as either High, Medium, or Low. According to experimental results, LightGBM outperforms Random Forest in terms of classification accuracy at 83.25%, and F1-Score at 0.8332 against 82.05% and 0.8106, respectively. Both algorithms have excellent prediction capabilities with ROC-AUC score above 0.93. Moreover, cross-validation tests confirm the accuracy, robustness, and consistency of the two models. In addition to their predictive prowess, both models demonstrate the importance of traditional academic variables like examination scores, assignments, and study habits. On the other hand, AI variables are secondary predictors in this case. This implies that AI has a positive impact on academic performance when combined with study habits.

Keywords: Artificial Intelligence in Education, Educational Data Mining, Student Performance Prediction, Machine Learning, Random Forest, LightGBM