Data-Driven Approach to Evaluate the Level of Service (LOS) of Demand-Responsive Transport for the Disabled (DRTD) with an ANFIS Algorithm

Abstract

Even though the Demand-Responsive Transport (DRT) service plays an important role to provide more convenience to the disabled people, they still have a challenge to be solved in terms of waiting time. Especially, the biggest problem is that waiting times for using the DRT service are variant across space and time because users cannot predict when they can get the service and/or how long they have to wait for service. This study developed a model for predicting the waiting time of Demand-Responsive Transport for Disabled (DRTD) with irregular spatiotemporal characteristics in real time. The primary purpose of the model developed was to monitor the level of service (LOS) to improve the satisfaction of users who use mobility services for the disabled and efficiently manage service providers. The model was estimated using an Adaptive Neuro-Fuzzy Inference System (ANFIS), which is known to have an excellent predictive performance by combining the advantages of both artificial neural networks and fuzzy inference systems. Four variables, including the number of calls (or requests), the number of vacant vehicles, Medical Infrastructure Concentration Index (MICI), and Disabled Population Concentration Index (DPCI), were used as input variables for the ANFIS-based model. Despite using cross-sectional data, the accuracy of the predicted model was found to be excellent (90%) and showed good and even prediction performance without bias by LOS categories The proposed method is expected to become a monitoring tool to manage mobility services and improve user convenience by notifying users how long they have to wait to use mobility services in real time.