A new binary genetic programming approach to designing public transportation systems according to transit-oriented development criteria

This study introduces a new evolutionary approach called Binary Genetic Programming (BGP) to design and assess public transportation systems from a sustainable development perspective. The BGP combines evolutionary system identi cation techniques with k-fold cross-validation to obtain an accurate model between the land use and transportation parameters from a sustainable urban development point of view. To assess the new model, two public transportation systems including the new tram line of Antalya (Turkey) and the bus rapid transit line of Bhopal (India) were considered. The model was employed to classify the transportation systems into Transit-Oriented Development (TOD) and non-TOD. The solutions generated by the new model were compared with those of classic Decision Tree (DT) as well as the state-of-the-art Random Forest (RF) models evolved as the benchmarks in this study. The results showed that the BGP was highly ethcient and might provide less than 5 classi cation error. It is superior to the DT and RF solutions, which typically require higher datasets to avoid over tting. Furthermore, the explicit formulation of BGP in combination with the multicriteria evaluation method increases human insight into the factors a ecting the design of public transportations from a sustainable urban development point of view.

Keyword: decision tree; genetic programming; random forest; transit-oriented development; urban planning