A Performance Evaluation Model for Taxi Cruising Path Recommendation System

Recommending an appropriate route to reduce taxi drivers’ mileage spent without a fare is a long-standing challenge. The current solution has been to get the best route which has optimal performance, and the performance usually combined the conditional probability for getting a passenger and the cruising distance. However, the main reference has some limitation. To eliminate the limitation, a novel model is proposed to evaluate the candidate route performance. And based on this new model, a recommendation system is tested. Firstly, by mining the knowledge of the historical taxi trajectory, we extract the temporal probabilistic recommending points. Then based on it, the evaluation model is presented to estimate the performance of each candidate route. Finally, a route recommendation algorithm is used to get the optimal route for taxi drivers. And as the result, the experiment is performed on real-world taxi trajectories data set, and shows the effectiveness of the proposed model for evaluating the performance.