Abstract
Passenger flow prediction is one of the most important issues in an urban subway system toward smart cities, which can help cut a trip time, plan a trip route, and thus provide a comfortable travel experience. However, there still exist some challenges on how to fully leverage effective and efficient passenger mobility patterns hidden in big traffic data to improve the accuracy of prediction. In this paper, we introduce a concept of the loyal passenger and propose a Spatio-Temporal Loyalty-based model (STLoyal) to improve the precision of prediction through analyzing the characteristics of loyal subway passengers. The proposed STLoyal model is evaluated using real-world subway transaction data sets in Shanghai, China, and it is compared with other state-of-the-art methods. The experimental results show that STLoyal yields superior prediction accuracy on weekdays and weekends in terms of loyalty, time, location, and weather metrics.