Predictive Modelling of Exam Outcomes Using Stress-Aware Learning from Wearable Biosignals

Area of Science:

• Educational Technology
• Biomedical Engineering
• Data Science

Background:

• Academic performance is influenced by various factors, including stress.
• Traditional methods for assessing academic performance and stress lack real-time insights.
• Wearable technology offers a non-invasive approach to collect continuous physiological data.

Purpose of the Study:

• To investigate the feasibility of using wearable technology and machine learning to predict academic performance.
• To examine the correlation between physiological stress indicators and academic outcomes.
• To develop a predictive model for student academic success based on physiological signals.

Main Methods:

• Collected six physiological signals (skin conductance, heart rate, skin temperature, electrodermal activity, blood volume pulse, inter-beat interval, accelerometer) using a wearable device during examinations.
• Developed a data preprocessing and feature engineering pipeline for physiological data.
• Trained and evaluated five machine learning models (Random Forest, SVM, XGBoost, CatBoost, GBM) using techniques like SMOTE, hyperparameter tuning, and dimensionality reduction.

Main Results:

• Physiological signals collected via wearable devices can effectively predict academic performance.
• A significant correlation was found between measured stress levels and academic outcomes.
• The developed machine learning models demonstrated the capability to forecast exam results based on physiological data.

Conclusions:

• Wearable technology combined with machine learning provides a viable method for predicting academic performance.
• Physiological stress monitoring can offer valuable insights into factors affecting student success.
• This approach holds promise for developing real-time monitoring systems to enhance student well-being and academic achievement.