해석가능한 기계학습을 활용한 보행목적별 보행만족도 영향요인 분석

초록

The uncoordinated development and expansion of cities along with industrialization and urbanization have caused various urban issues, such as traffic congestion, heavy energy consumption, and environmental pollution. To alleviate the negative impacts of the automobile-centered urban environment, pedestrian-oriented urban planning and design practices have been proposed and implemented for several decades in Seoul, Korea. However, the factors influencing the walking satisfaction by walking purpose have not been sufficiently investigated in urban planning and design literature. In addition, the majority of previous studies include attempts to diagnose only the linear relationships between the walking satisfaction and the built environment. With the recent development of an interpretable machine learning (IML) model, the nonlinear relationships between the walking satisfaction and the built environment is investigated in this study. Further, the interaction effects of the built environmental variables on the walking satisfaction are identified. The results indicate that the machine learning model shows a significantly higher explanatory power compared with the conventional model. In addition, it is confirmed that IML is a useful tool to understand the nonlinear relationships between the walking satisfaction and the built environment. The analysis results on the aforementioned relationship suggest that they can be used as important data to promote a pedestrian-friendly urban environment. Additionally, the policy implications of promoting walking satisfaction by purpose are presented in this study.

키워드