AI-Driven Student Performance Prediction Using Behavioral, Academic and Lifestyle Analytics

B. Rithuja Sai Sri, R. S. Jessie Bernice David, B. Sowmya, R. Regin, K. Senthamilselvan, Prasanna Ranjith Christodoss Department of Computer Science and Engineering in Artificial Intelligence and Machine Learning, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India. Department of Computer Science and Engineering, SRM Institute of Science and Technology, Ramapuram, Chennai, Tamil Nadu, India. Department of Electronics and Communication Engineering, Dhaanish Ahmed College of Engineering, Chennai, Tamil Nadu, India. Department of Computing, Mathematics and Physics, Messiah University, Mechanicsburg, Pennsylvania, United States of America.

Abstract: Modern education relies on student success prediction due to personalised learning and early academic support. Most traditional evaluation methods rely on exam scores, which don't necessarily reflect student learning or its effects. This research uses AI to link academic data with attendance, study habits, and sleep length. Researchers found that adding efficiency, balance, and fatigue to student behavior better represents student behavior than raw input alone. Researchers started with Decision Trees, SVMs, and Random Forests. Only Random Forest produced more accurate results and performed better with multiple inputs; thus, researchers preferred it. Balanced study, regular practice, and decent sleep were also prioritised for student performance. This was our final model. A synthetic dataset was created to imitate student study habits and daily routines without student data. The dataset includes self-study, practice, revision, project work, attendance, sleep length, efficiency, balance, and burnout. The model was 90% accurate after an 80–20 data split for training and testing. The F1 score, precision, and recall helped us evaluate it. A balanced study distribution, frequent practice and review, and sufficient sleep affected expected outcomes in our trials. The program anticipates and provides pupils with simple, personalised study advice. Python model data was processed with numpy and pandas. Sckitlearn for machine learning and Streamlit for user-friendly interaction. This study implies that machine learning and behavioral insights can improve student performance analysis.

---

Keywords: Student Performance Prediction; Random Forest; Learning Analytics; Educational Data Mining; Behavioral Analytics; Feature Engineering; Predictive Modeling.

---

Received on: 28/05/2025Revised on: 15/07/2025Accepted on: 14/08/2025Published on: 01/03/2026

---

---