Improved Stress Prediction Using Differential Boosting Particle Swarm Optimization-Based Support Vector Machine Classifier

Abstract

The emerging technology and stress-causing events in the life span of humans impose an effective stress prediction system. The intervention of early detection of stress will effectively reduce the stress and enriches the quality of life. With the assistance of a machine learning approach, a new stress prediction model is developed and it gives personalized as well as an adaptive stress prediction model. The process of learning uses physiological signals, which effectively identifies the stress status of the user. The identification and selection of the best features is a vital step of data preprocessing in case of dimensionality reduction. The performance of the classification model is degraded when the model is trained without reducing the dimensionality of the dataset which may result in poor performance. Hence, identifying and selecting the best features can improve the performance of the classifiers. The proposed stress prediction model with Differential Boosting Particle Swarm Optimization retrieves the best features and the Support Vector Machine (SVM) classifies the subjects into three categories, namely, stress, normal, and relax. The proposed DBPSO-based SVM is compared with the existing approaches and it is evaluated using six performance metrics. From the experimental results, the proposed model attains high accuracy with a low classification error rate.