Innovative computation to detect stress in working people based on mode of commute

Abstract

Introduction:

Commuting is an integral part of modern life for many people, shaping daily routines and impacting overall well-being. With various transportation options, including driving, public transport, walking, and cycling, commuters encounter various experiences and challenges in their everyday journeys. Understanding how different modes of commuting affect stress levels is essential for improving public health and informing transportation planning. This study develops advanced machine-learning techniques to explore the connection between commuting methods and stress levels.

Methods:

This research examines how different commuting modes affect stress levels using machine learning methods. The study analyses data collected from 45 individuals who regularly commute to work, focusing on driving, cycling, and public transport modes. Non-invasive wearable sensors were utilised to gather electroencephalography (EEG), blood pressure (BP), and heart rate (HR) data for five consecutive days for each participant. Additionally, qualitative data was collected using the Positive and Negative Affect Schedule (PANAS) questionnaire to assess participants’ emotional responses before and after their commute. The research focuses on developing a machine learning-based model to predict the commute’s impact and monitor the stress level due to the commute mode. In research, objective and subjective factors shape the research process and outcomes. Understanding the interaction between these factors is essential for conducting thorough and reliable research that produces valid results. Our study utilises datasets incorporating qualitative and quantitative data from questionnaires and human bio-signals.

Results:

This research developed various machine learning algorithms to detect stress levels based on commuting mode. The results indicate that the Linear Discriminant Analysis technique achieved an accuracy of 88%, while Logistic Regression reached 90.66% accuracy. The Boosted Tree algorithm produced the best performance, with an accuracy of 91.11%. Furthermore, incorporating personalised parameters into the data improved the accuracy of these algorithms in detecting stress levels. Cross-validation was also utilised to mitigate the risk of overfitting, ensuring robust and reliable model performance.

Conclusion:

The findings reveal that human bio-signals tend to increase following commuting, irrespective of the mode, with driving identified as the most stressful option. Commuters using passive modes of transport experience elevated stress levels compared to those using active modes. This research underscores the importance of understanding the connection between commuting modes and stress, providing key insights into the potential health impacts of daily travel. The development of an intelligent model to predict stress levels based on commuting mode offers valuable contributions to public health and transportation planning, with the goal of enhancing well-being and improving commuters’ quality of life.