Scholarlink

Stress detection is critical for mental and physical well-being, highlighting the need for efficient machine learning solutions. This research develops boosting models such as Extreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), Categorical Boosting (CatBoost) and Gradient Boosting Machine (GBM), optimized using Bayesian hyperparameter tuning for enhanced stress classification. A dataset of 320 individuals, including physiological and activity-based parameters (heart rate, temperature, humidity, step count and blood pressure), was collected for model training. The proposed model achieves a maximum accuracy of 99% (XGBoost), outperforming GBM (98%), CatBoost (96%), and AdaBoost (75%). SHAP analysis was conducted to interpret feature importance, revealing body temperature, step count, and blood pressure as key indicators of stress. A comparative analysis with state-of-the-art methods confirms the model’s robustness, high accuracy, and real-world applicability for stress monitoring and intervention.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhancing Stress Detection Using Bayesian Optimized Boosting Algorithms and Feature Analysis Using SHAP

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Mohammad Rabib Uddin,
  • Md. Saniat Rahman Zishan,
  • Shameem Ahmad,
  • Upama Dev

摘要

Stress detection is critical for mental and physical well-being, highlighting the need for efficient machine learning solutions. This research develops boosting models such as Extreme Gradient Boosting (XGBoost), Adaptive Boosting (AdaBoost), Categorical Boosting (CatBoost) and Gradient Boosting Machine (GBM), optimized using Bayesian hyperparameter tuning for enhanced stress classification. A dataset of 320 individuals, including physiological and activity-based parameters (heart rate, temperature, humidity, step count and blood pressure), was collected for model training. The proposed model achieves a maximum accuracy of 99% (XGBoost), outperforming GBM (98%), CatBoost (96%), and AdaBoost (75%). SHAP analysis was conducted to interpret feature importance, revealing body temperature, step count, and blood pressure as key indicators of stress. A comparative analysis with state-of-the-art methods confirms the model’s robustness, high accuracy, and real-world applicability for stress monitoring and intervention.