Personalized service recommendations for travel using trajectory pattern discovery

Service recommendations help travelers locate en route traffic information service of interest in a timely manner. However, recommendations based on simple traffic information, such as the number of requests for the location of a facility, fail to consider an individual’s preferences. Most existing work on improving service recommendations has continued to utilize the same ratings and rankings of services without consideration of diverse users’ demands. The challenge remains to push forward the modeling of spatiotemporal trajectories to improve service recommendations. In this research, we proposed a new method to address the above challenge. We developed a personalized service-trajectory correlation that could recommend the most appropriate services to users. In addition, we proposed the use of “congeniality” probability to measure the service demand similarity of two travelers based on their service-visiting behaviors and preferences. We employed a clustering-based scheme, taking into account the spatiotemporal dimensions to refine the trajectories at each spot where travelers stayed at a certain point in time. Experiments were conducted employing a real global positioning system–based dataset. The test results demonstrated that our proposed approach could reduce the deviation of the trajectory measurement to 10% and enhance the success rates of the service recommendations to 60%.