Intelligent experimental data processing and analysis for academic performance prediction using mobile phone usage data

The extensive use of smartphones by university students has resulted in the continuous generation of behavioral data that offers valuable insights into learning habits and academic outcomes. This study proposes an intelligent experimental framework for the processing and analysis of mobile phone usage data aimed at predicting academic performance. The framework combines advanced data preprocessing methods, robust feature engineering techniques, and both machine learning and deep learning models to identify meaningful behavioral patterns associated with student achievement. Mobile phone usage records including application usage behavior, screen interaction duration, temporal activity patterns, and usage intensity were collected throughout an academic semester and systematically aligned with students’ academic performance indicators. To address data incompleteness and noise, a behavior-aware imputation strategy was employed, followed by normalization, feature selection, and dimensionality reduction to improve model stability and generalization. Multiple predictive models, including Random Forest, Support Vector Machine, Multilayer Perceptron, Long Short-Term Memory, and Bidirectional Long Short-Term Memory networks, were evaluated using established performance metrics. The experimental findings indicate that deep learning models, particularly the Bi-LSTM architecture, consistently outperform traditional machine learning approaches by effectively modeling temporal dependencies in mobile usage behavior. These results underscore the potential of intelligent mobile data analytics as a reliable tool for academic performance prediction and support its application in data-driven educational monitoring and early intervention systems.