Skip to content
Conference Open access

Feature Selection and Explainable AI for Stress Level Prediction using Multi-Modal Behavioral Data

Sep 2026 · 2026 10th International Conference on I-SMAC (IoT in Social, Mobile, Analytics and Cloud) (I-SMAC) · pp. 2012-2017 · 0 citations · 12 references

Abstract

Psychosocial stress is an important global public health concern and has a significant impact on emotional, physical, academic and work-related productivity. This study presents an Explainable Artificial Intelligence (XAI) framework for predicting stress level based on psychological, behavioral, physiological, sleep and mobility features. The publicly available dataset used in this study consisted of 3000 observations from 100 participants, who were monitored over a 30-day period and were monitored for stress. Z-score normalization was used for data preprocessing and Random Forest feature importance was used to determine which predictors are the most informative to be used. Stratified 5-fold cross validation with hyperparameter optimization using GridSearchCV was used to train an Extreme Gradient Boosting (XGBoost) classifier. A novel approach to achieving model interpretability was developed by employing SHAP (SHapley Additive exPlanations) to measure each feature's contribution to the prediction of stress. Experimental results showed good predictive performance with 93.20% accuracy, 92.90% macro-precision, 92.50% macro-recall and 92.70% macro-F1-score. The most significant predictors were skin conductance, mobility radius, sleep duration, sleep time, and accelerometer. SHAP analysis also highlighted the direction and magnitude of feature contributions, providing additional transparency and trust in the model. The results show that the proposed framework is a viable and explainable approach for stress prediction based on multi-modal behavioral signals obtained and integrated through feature selection, XGBoost classification and explainable artificial intelligence.

Read PDF

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.