Mining GPS Data to Learn Driver's Route Patterns

Abstract

Over the last few years, GPS guidance systems have become increasingly more popular. GPS-equipped devices like smart phones become more common and larger amounts of GPS data become available to geographic applications. Having precise information about the routes of a driver during a period of time can be useful to learn and estimate both the traffic and the driver's intent at specific moment of time. With our solution we want to go a step further to the existing GPS navigation systems by designing a mechanism that is capable to learn driver's routes. We could offer in the future a point-to-point concept for an environmentally friendly routing mechanism anywhere within a selected road network based on our HMM-method and a training process. Our study is based on real data collected from various local drivers and can be easily applied in modern intelligent traffic systems. The system comes with a user interface that displays the GPS routes on the map for a specific driver. These routes can be analyzed using parameters like time, distance, height and speed. Also we developed a tool that manages to compute the maximum-likelihood using the Viterbi algorithm in order to validate the next route segment election for a sampled road network.