Optimizing operational parameters through minimization of running costs for shared mobility public transit service: an application of decision tree models

The aim of this study was to use machine learning model for prediction of running costs of public transport buses in Karachi, which is the most widely used mode of shared mobility in this city. To achieve this objective, classification and regression tree (CART) models have been used. Later on, the prediction models were used to determine the optimum operating parameters of public transport buses in Karachi. An interview study was conducted to acquire their operational and cost parameters. The dataset comprised for operational and maintenance parameters of 146 buses of various specifications. Running costs were calculated on the basis of number of passengers, hours of service and distance traveled. From the CART models, it was found that the minimum total weekly distance traveled by the vehicle should be 297 km, below which costs start to increase. Similarly, optimum values could also be found, from CART models, for other parameters, such as; number of passengers, number of hours in which the vehicle was running, and type of vehicle. The findings of this study would be helpful in designing future shared mobility public transit options, such as metro.