A Comparative Study of Classification of Occupational Stress in the Insurance Sector Using Machine Learning and Filter Feature Selection Techniques

Abstract

In recent years, occupational stress mining has become a widely exciting issue in the research field. The primary purpose of this study is to analyze filter feature selection methods for the efficient occupational stress classification model. We propose and examine seven different techniques of filter feature selection such as Chi-Square, Information Gain, Information Gain Ratio, Correlation, Principal Component Analysis, and Relief. The resultant selected features are then used with popular classifiers like Naive Bayes (NB), Random Forest (RF), Support Vector Machine (SVM), Artificial Neural Network (ANN), and Gradient Boosted Trees (GBT) for detection of occupational stress in the insurance sector. A survey-based psychological primary occupational stress data set is used to evaluate the relative performance of these methods. This study effectively demonstrated the significance of filter feature selection methods and explained how accurately they could help classify stress levels. This study showed that the Correlation-based feature selection with the SVM classifier obtained the best performance compared to other filter feature selection methods and classification models.