Prediction of walking occurrence and frequency among the residents of city community housing projects using machine learning techniques

Introduction Walking is one of the most affordable and accessible forms of daily physical activity, yet it remains an unsafe and under-supported mode of transport in many low-income urban neighborhoods. Residents of public housing communities often face inadequate pedestrian infrastructure, traffic safety risks, and limited access to essential services, all of which discourage walking and exacerbate health inequalities. Understanding the complex interplay of these factors is critical for improving mobility and public health outcomes among vulnerable urban populations. Methods This study applies a Bayesian Network model to predict the occurrence and frequency of walking among 680 residents across five low-income public housing projects (PPRs) in Kuala Lumpur, Malaysia. Using cross-sectional data on individual demographics, travel behavior, built environment perceptions, and health status, the model identifies probabilistic relationships between key factors influencing pedestrian activity. Results Results indicate that gender, employment status, and perceived walkability significantly affect the likelihood and frequency of walking trips. Notably, the perceived quality of pedestrian infrastructure and accessibility to public facilities also emerged as influential predictors. Conclusions The results highlight the urgent need to design safer and more inclusive walking environments in low-income urban areas. The study demonstrates the practical utility of Bayesian modeling in capturing interdependent factors influencing walking behavior and provides actionable insights for planners and policymakers seeking to promote equitable, health-oriented, and pedestrian-friendly urban mobility.