Predicting the Upcoming Services of Vacant Taxis near Fixed Locations Using Taxi Trajectories
AbstractEmerging on-line reservation services and special car services have greatly affected the development of the taxi industry. Surprisingly, taking a taxi is still a significant problem in many large cities. In this paper, we present an effective solution based on the Hidden Markov Model to predict the upcoming services of vacant taxis that appear at some fixed locations and at specific times. The model introduces a weighted confusion matrix and a modified Viterbi algorithm, combining the factors of time of day and traffic conditions. In our framework, the hotspot or hidden states extraction is implemented through kernel density estimation (KDE) and fuzzy partitioning of traffic zones is done via a Fuzzy C Means (FCM) algorithm. We implement the proposed model on a large-scale dataset of taxi trajectories in Beijing. In this use case, tests demonstrate the high accuracy of the modeling framework in predicting the upcoming services of vacant taxis. We further analyze the factors affecting the predictive accuracy via a prediction accuracy analysis and prediction location evaluation. The findings of this paper can provide intelligence for the improvement of taxi services, to increase the passenger capacity of taxis and also to improve the probability of passengers finding taxis. View Full-Text
Share & Cite This Article
Hu, C.; Thill, J.-C. Predicting the Upcoming Services of Vacant Taxis near Fixed Locations Using Taxi Trajectories. ISPRS Int. J. Geo-Inf. 2019, 8, 295.
Hu C, Thill J-C. Predicting the Upcoming Services of Vacant Taxis near Fixed Locations Using Taxi Trajectories. ISPRS International Journal of Geo-Information. 2019; 8(7):295.Chicago/Turabian Style
Hu, Chunchun; Thill, Jean-Claude. 2019. "Predicting the Upcoming Services of Vacant Taxis near Fixed Locations Using Taxi Trajectories." ISPRS Int. J. Geo-Inf. 8, no. 7: 295.
Note that from the first issue of 2016, MDPI journals use article numbers instead of page numbers. See further details here.