Capturing impressions of pedestrian landscapes used for healing purposes with decision tree learning

In 2002, Medicare health insurance recognized the relationship between pedestrian environments and public health by providing co-pay for health care delivered in residential land use. In this multi-disciplinary experiment, artificial intelligence (AI) and landscape architecture (LA) bridge their respective domains to measure and model the pedestrian reaction to walking environments in small town residential community in Central Texas. In the process, we gained a deeper understanding of the health motivation of walkers and their empirical relationship to various street environments that they used for health purposes. The analytical model we ultimately developed is a flexible tool that facilitates exploration of people's perceptions of the landscape, how the pedestrian landscapes are functioning in the opinion of its users, and how changes to the design of the walking domain may predictably affect physical activity levels with the associated health benefits. A pilot study involving fifty-four participants and six walking environments were used in the development of an analytical model that is significantly site-specific and grass roots oriented. Participant perceptions were measured querying each participant's rating of fifty discrete environmental variables taken. This data was then analyzed using the decision tree algorithm. Our primary objective was to capture the decision-making pattern walkers engage in when deciding to walk in a particular environment specifically for health purposes and to make this available to the designers of pedestrian environments in transportation corridors. The approach gave the designers new insight into the critical variables and the not so critical variables that affected people's decision to walk for health purposes. The results from the analysis defined measurable environmental variables that form the design for pedestrian activity in the six walking environments in the study area. A customized version of decision tree machine learning algorithm rules for designing good pedestrian landscapes for health purposes were extracted from the grass roots surveys. The data indicated that variables influencing the decision to walk for health purposes in the study area included weather, sound, water, light and edge of space. The analytical model derived from the discipline of artificial intelligence facilitated examining a subset of variables and manipulating of individual or group of these variables to better understand how the built environment affected decisions to walk for different purposes. This collaboration was our first phase in developing intelligent tools for designers that provided site-specific user-specific data to the planner or designer of pedestrian space.