Predict Personal Continuous Route

Abstract

In the daily life, people often repeat regular routes in certain periods. Predicting personal future routes using this information helps to achieve many goals, including improving the quality of intelligent transportation systems (ITSs) and location-based services (LBSs) for individuals. In this paper, a novel system is developed to predict the personal future routes based on the continuous route patterns extracted in advance. The proposed approach predicts a person's future route through the use of a probabilistic tree model built from his / her route patterns. The route patterns are extracted from personal history of movement using a new mining algorithm, continuous route pattern mining (CRPM), which based on PrefixSpan. Furthermore, the separated system architecture guarantees the safety of personal privacy while greatly reducing the computational load on mobile devices. An evaluation using a corpus of real routes from 17 persons demonstrates the effectiveness of the system. Using only a month recorded trips data, our system can get an average correct rate of about 74.3% in one step predicting. In route prediction, the average Levenshtein distance between the real trips and predicting results produced by our system is about 30% shorter than that produced by the basic Markov method.