Statistical analysis based on a two-staged SEM-SVM approach for environmental noise annoyance prediction and identification of influencing factors

BACKGROUND AND OBJECTIVES: Noise pollution is an environmental stressor that is mainly caused due to heavy transportation in urban scenarios. Traffic noise is a growing concern in urban environments, impacting public health and well-being. As urbanization expands, understanding and mitigating traffic-induced noise annoyance becomes increasingly critical. This study aimed to develop a machine-learning model for predicting traffic-induced noise annoyance in Riyadh, Saudi Arabia. The research explored the influence of factors like demographics, noise characteristics, and traffic conditions on noise annoyance.METHODS: A survey was conducted at 21 locations in Riyadh, collecting data from 928 participants. The survey included questions on demographics (gender, age, education, marital status, profession), traffic conditions (traffic flow), and noise perception (transportation noise, noise sensitivity, perceived noisiness). The sampling method employed was a combination of stratified and random sampling. Stratified sampling was used to ensure that various demographic segments (e.g., different age groups, genders, and education levels) were proportionately represented in the survey. Structural Equation Modeling was used to analyze the collected data and identify factors significantly affecting noise annoyance. These significant factors were then used as input variables for a Support Vector Machine model designed to predict noise annoyance. The performance of the Support Vector Machine model was evaluated using Root Mean Square Error, Mean Absolute Error, and R-squared.FINDINGS: The Structural Equation Model analysis revealed that gender, age, education level, traffic flow, noise from traffic, and individual noise sensitivity were significant contributors to noise annoyance. The developed Support Vector Machine model achieved a high level of accuracy with a root mean square error of 1.416 and a coefficient of determination of 0.90. Noise sensitivity emerged as the most crucial factor influencing noise annoyance.CONCLUSION: This study demonstrates the effectiveness of machine learning, specifically the Support Vector Machine, in predicting traffic-induced noise annoyance. The findings highlight the importance of both individual characteristics and environmental factors in noise perception and can be valuable for urban planning and noise mitigation strategies, promoting a more noise-resilient city environment. For the community, urban planners and policymakers can use these findings to design silent areas by implementing noise barriers, optimizing traffic flow, and enforcing stricter noise regulations.