IDENTIFYING PSYCHOLOGICAL HEALTH STATUS OF EMPLOYEES OF AN ORGANIZATION BY USING FEATURE SELECTION AND MACHINE LEARNING TECHNIQUES

Abstract

Mental health has always been a major concern across the globe, but when it comes to the mental health of employees one way or another it is somehow neglected. Due to workload and short deadlines employees have always been under a great burden and thus more prone to mental health disorders. Poor mental health leads to degradation in employees performance and productivity. So in this project, we proposed a model to detect the mental health state of organization employees, and determine the key features affecting the mental health of employees using feature selection. We used randomized Hyperopt in conjunction with Xgboost to predict the mental health state of employees. Data used belongs to the year 2019 and 2020, where the year 2020 refers to the mental health state of employees during the coronavirus pandemic. A total of three datasets has been used to determine the mental health state of employees before and during novel coronavirus and a comparison is made to check the increase in the number of employees with mental health disorders during coronavirus. Feature selection is performed to determine the features affecting the mental health of employees during coronavirus. Randomized Hyperopt and Xgboost were found more effective in predicting the mental health state of employees than traditional machine learning algorithms during coronavirus. It was found to have an accuracy of 91% when predicting mental health disorders. In the 2019 dataset mental disorders are used to determine the mental health state of employees, where having a mental disorder refers to bad mental health. Whereas in the 2020 dataset or during covid 19 lockdown mental health fatigue rate of employees is used as a baseline to predict the mental health state of employees. A comparison with some machine learning techniques to determine how effectively the mental health state of employees is predicted.