Computer based pedestrian landscape design using decision tree templates

Machine Learning algorithms can act as a valuable analytical tool in design research. In this paper, we demonstrate the application of a decision tree learning algorithm for designing pedestrian landscapes that encourage walking for health. The domain knowledge was captured using intercept surveys that queried responses to cognitive, physical and social attributes that influence pedestrian spatial analysis. Decision trees extracted from the knowledge base were used in the design of pedestrian landscapes, which were tested in a transportation simulator. The observed match between the change in the participants' response to manipulation of physical variables in the simulated world with those predicted by the decision rules indicate the appropriateness of applying decision tree rules as guidelines during the process of pedestrian landscape design and research.