Personal Continuous Routes Prediction Based On Route Patterns Mining

Recently, with the development of positioning and popularity of pervasive computing, the trajectories and positions related services and products became the focus of industry. For example, the modern Intelligent Transport System (ITS) utilized the sensor which equipped on vehicles to get information for transport scheduling improvement. Another example is the Location-Based Services (LBS) which get data of vehicles from portable device and make use of it to provide real-time services for the drivers. In these fields, route patterns mining and predicting for moving objects is the foundation.Until now, the major object in the route and position related research is the vehicle, not the person. However, in daily life, consider a person as a moving unit is more pervasive than consider a vehicle as a moving unit. In this paper, we proposed an entire system for personal continuous routes mining and predicting. The System takes into consideration the following issues: the diversity of personal moving status, the precision of the positioning system, the limited computing ability of personal portable device, the privacy of personal routes and the reusability of the modules in the system.To handle these issues, we proposed several novel solutions. First, mobile phone and mini GPS sensor were used for data collection and processing. A program which ran on mobile phone can record the user's moving status adaptively. Several data filters are used for cleaning data and remove the noise. Second, we proposed the Continuous Route Patterns Mining (CRPM) which can tolerate the practical disturbance in real route data and extract long route patterns. Third, we designed a new Client/Server architecture. This architecture can make the server conducts most of computation without get awareness of users' real routes, which will greatly reduce the computational load on the personal portable devices. Fourth, we proposed a Decision-Tree based route predicting algorithm, which can not only reduce the computation, but also provide high predicting precision. An Evaluation over 17 users show that, CRPM can extract much longer route patterns than the traditional sequential mining algorithms, and our predicting algorithm can provide real time and effective solution for personal route prediction.