Ridesharing car detection by transfer learning

Abstract

Ridesharing platforms like Uber and Didi are getting more and more popular around the world. However, unauthorized ridesharing activities taking advantages of the sharing economy can greatly impair the healthy development of this emerging industry. As the first step to regulate on-demand ride services and eliminate black market, we design a method to detect ridesharing cars from a pool of cars based on their trajectories. Since licensed ridesharing car traces are not openly available and may be completely missing in some cities due to legal issues, we turn to transferring knowledge from public transport open data, i.e., taxis and buses, to ridesharing detection among ordinary vehicles. We propose a novel two-stage transfer learning framework, called CoTrans. In Stage 1, we take taxi and bus data as input to learn a random forest (RF) classifier using trajectory features shared by taxis/buses and ridesharing/other cars. Then, we use the RF to label all the candidate cars. In Stage 2, leveraging the subset of high confident labels from the previous stage as input, we further learn a convolutional neural network (CNN) classifier for ridesharing detection, and iteratively refine the RF and CNN, as well as the feature set, via a co-training process. Finally, we use the resulting ensemble of the RF and CNN to identify the ridesharing cars in the candidate pool. Experiments on real car, taxi and bus traces show that CoTrans, with no need of a pre-labeled ridesharing dataset, can outperform state-of-the-art transfer learning methods with an accuracy comparable to human labeling.