Classification of Stress in Office Work Activities Using Extreme Learning Machine Algorithm and One-way ANOVA F-Test Feature Selection

Abstract

Stress is a condition when humans feel psychological pressure consciously or unconsciously comes from within themselves and the environment. Based on the survey, more than 50% of workers experienced stress at work. Neglecting stressful conditions repeatedly can worsen the performance and health of workers. Identifying stress condition manually is not effective and may take some time. There is a need to build a system to diagnose stress quickly and accurately. Machine learning classification is one of the solutions to problems that are applied to intelligent systems. One of the machine learning methods that can be used is the Extreme Learning Machine (ELM) algorithm. One-way ANOVA F-test is used as a method of feature selection to require a quick decision so the reduction in features is expected to accelerate the results of the classification. The dataset used is the Heart Rate Variability totaling 5000 samples with 35 features and 3 classes. Based on experiments, the conventional ELM algorithm produces an accuracy of 0.878 while the combination of the ELM algorithm and the One-way ANOVA F-test produces an accuracy of 0.91 with 33 selected features. Thus, the effect of using the feature selection method can increase accuracy and reduce computational time, and the addition of hidden neurons results in a significant increase in accuracy and computational time.