Routing quality always dominates the top 20% of in vehicle- navigation customer complaints. In vehicle navigation routing engines do not customize results based on customer behavior. For example, some users prefer the quickest route while some prefer direct routes. This is because in vehicle navigation systems are traditionally embedded systems. Toyota announced that new model vehicles in JP, CN, US will be connected with routing function switching from the embedded device to the cloud in which there are plenty of probe data uploaded from the vehicles. Probe data makes it possible to analyze user preferences and customize routing profile for users. This paper describes a method to analyze the user preferences from the probe data uploaded to the cloud. The method includes data collection, the analysis model of route scoring and user profiling.
Furthermore, the evaluation of the model will be introduced at the end of the paper. The analysis not only focuses on the routes chosen by the user but also compares with the ones not chosen for the same ODs using multivariate analysis on the preference of each user with the routing parameters such as time, toll, etc. Moreover, in the paper, it will introduce the method to project the preference of each user to the entire distribution composed of all users to estimate the preference including the ODs that users have not experienced.