Abstract
Stress is major health related issue globally. It is important to manage the stress for the healthcare as well as the quality of life of an individual. For this purpose an effective classification model is required. Therefore, a novel stress classification model based on convolutional neural network (CNN) has been proposed. The aim of this paper is to enhance the accuracy of stress prediction model using CNN for a participant in a real-world environment based on insurance sector data. To obtain the desired model the data set cleaned and statistically analyzed using z-score, Annova test, paired t-test and bootstrap paired t-test. On the basis of this analysis stress detection featured identified. We performed an experiment using the dataset by applying CNN. For assessing the performance of proposed classifier we made a comparison with other stress prediction model. The OSI questionnaire used for primary data collection. It has total nine demographic variables and 12 sub scales for assessing occupational stress. The questionnaire has total 46 items related to these 12 dimensions, and each one is rated on 5 point Lickert scale. The northern Indian states were surveyed for data collection. Out of 630 participants 500 were finally selected. They were categorized into gender age, education, management, work experience, marital status and religion. The implementation performed using R open source environment and SPSS. In this proposed CNN model SGD optimization technique employed for reducing the error and enhancing the accuracy. The statistically analyzed dataset was used as an input for deep learning algorithm called convolution Neural Network for stress classification modeling. In this analysis, 280 epochs and approximately 500 Instances used. The experimental result obtained from the proposed model succeeded to achieve an accuracy of 75.52 and specificity of 80.42. The performance of proposed CNN model is found better in comparison to artificial neural network. The proposed model is able to effectively classify the stress and it can be applied for stress prediction in stress monitoring system