Research On Users' Trajectory Mining And Prediction Based On GPS Data

In recent years, with the popularity of mobile Internet technology and the rapid development of intelligent mobile terminals, More and more variety of rich and colorful intelligent mobile applications and services are being generated and gradually becoming a part of people's daily life. At the same time, these mobile applications and services produced a huge amount of location-related data and interactive business data. How to mine these massive data has become the focus of the current research.The mining of the users' trajectory data and the research of user behavior prediction based on the GPS, has many applied scenarios. For an individual user, through the mining of the user' historical trajectory data,machine learning can learn the user's personal information such as interest,occupation, habits, and then provide the user with personalized push service, such as intelligent travel guide, position remind services, music recommendation, accurate advertising push, etc. For the group of the users,through the model of studying all users' trajectory data, machine learning can learn regional hot spots, hot activity and the relationships between different users, etc. It has great applied value of the intelligent city planning,intelligent transportation and popular activity monitoring.Most of the existing location service directly using the data provided by the user, lacking the deep mining of the research on user trajectory data,unable to learn a single user's behavior habit and groups of users' moving flow and provide targeted personalized service. This paper, based on the user's mobile trajectory data and geographic information data, make an in-depth study of how to learn a single user's or groups' places of interest,how to extract and rectify the user's moving path, how to model the users'trajectory data and predict people's behavior. Through studying of domestic and foreign research results, this paper put forward to the places of interest's clustering algorithm based on density clustering algorithm, the path extraction and correction algorithm, the users' behavior prediction algorithm based on BP neural network model. Through the experimental analysis, the algorithms proposed in this paper have a good reliability,accuracy and innovation.