Title: Mining Public Transit Ridership Flow and Origin-Destination Information from Wi-Fi and Bluetooth Sensing Data

Abstract: Transit ridership flow and origin-destination (O-D) information is essential for enhancing transit network design, optimizing transit route and improving service. The effectiveness and preciseness of the traditional survey-based and smart card data-driven method for O-D information inference have multiple disadvantages due to the insufficient sample, the high time and energy cost, and the lack of inferring results validation. By considering the ubiquity of smart mobile devices in the world, several methods were developed for estimating the transit ridership flow from Wi-Fi and Bluetooth sensing data by filtering out the non-passenger MAC addresses based on the predefined thresholds. However, the accuracy of the filtering methods is still questionable for the indeterminate threshold values and the lack of quantitative results validation. By combining the consideration of the assumed overlapped feature space of passenger and non-passenger with the above concerns, a three steps data-driven method for estimating transit ridership flow and O-D information from Wi-Fi and Bluetooth sensing data is proposed in this paper. The observed ridership flow is used as ground truth for calculating the performance measurements. According to the results, the proposed approach outperformed all selected baseline models and existing filtering methods. The findings of this study can help to provide real-time and precise transit ridership flow and O-D information for supporting transit vehicle management and the quality of service enhancement.