Enhanced Ensemble Fusion Model for Stress Classification and Prediction

Abstract

Stress has become a common phenomenon in modern society, and it has been identified as a major factor that affects people's health and well-being. Stress can be caused by various factors, such as work pressure, financial difficulties, relationship problems, and health issues. Prolonged exposure to stress can lead to physical and mental health problems, including anxiety, depression, cardiovascular diseases, and obesity. Accurate stress classification and prediction can help individuals and organizations identify the sources and levels of stress and take appropriate measures to manage stress and prevent negative outcomes. By identifying individuals who are at risk of stress, proactive interventions can be initiated to prevent negative outcomes. Additionally, stress classification and prediction can be useful for designing effective stress management programs and policies that can improve the well-being and productivity of individuals and organizations. Existing systems for stress classification and prediction have limitations in terms of accuracy and efficiency. To overcome these limitations, this paper proposes an Enhanced Ensemble Fusion (EEF) model that combines three ensemble classifiers, namely stacking, bagging, and boosting, using a blending classifier. The EEF model is composed of several classifiers, including the stacking classifier, the bagging classifier, and the boosting classifier, each using an Enhanced J48, Enhanced SVM, and Enhanced Naive Bayes classifier. An Enhanced Logistic Regression classifier is used as a meta-classifier for the stacking classifier. The model was evaluated on a Swell-EDA dataset and WESAD-EDA dataset, and the results show that it outperformed existing systems in terms of accuracy and robustness. The Enhanced Ensemble Fusion Model achieved an accuracy of 72.86% for WESAD-EDA dataset and 50% for Swell-EDA dataset which is significantly higher than the accuracy of individual classifiers and existing ensemble methods. The proposed model provides a promising approach for stress classification and prediction, which can be useful in various applications, such as healthcare, human resources, and education.