Research on User Behavior Prediction and Profiling Method Based on Trajectory Information
- PDF / 889,808 Bytes
- 10 Pages / 612 x 792 pts (letter) Page_size
- 87 Downloads / 231 Views
esearch on User Behavior Prediction and Profiling Method Based on Trajectory Information Hao Lia and Haiyan Kangb, * a
Computer School, Beijing Information Science and Technology University, Beijing, 100192 China School of Information Management, Beijing Information Science and Technology University, Beijing, 100192 China *e-mail: [email protected]
b
Received December 4, 2019; revised February 17, 2020; accepted February 19, 2020
Abstract—Aiming at the need to discover user behavior characteristics and knowledge from moving trajectory data, a user behavior profiling method based on moving trajectory information was proposed. Firstly, the trajectory coordinates were preprocessed to clean out good data. Secondly, the travel rules and the points of interest of the user were found by means of stay points detection, staying points’ semantics and frequent pattern mining. In the aspect of predicting user trajectory information, Key Points Long Short-Term Memory Networks (KP-LSTM) was proposed to predict the user’s future travel location; then the user’s important attribute characteristics were taken through the user profiling, intuitively depicting the characteristics and patterns of users’ lives. Finally, the availability of the method was proved by experiments, and the prediction accuracy was better than the traditional Linear regression and LSTM neural network. Keywords: trajectory data mining, user profile, LSTM, behavioral analysis, trajectory visualization DOI: 10.3103/S0146411620050065
1. INTRODUCTION The precise positioning, dynamic update of the communication service used by the user, and the massive acquisition of the user’s location information provide data support for fine-grained analysis of the user’s travel habits. The analysis and mining of these data can deeply understand the user behavior pattern, and manage the mobile object, which provides a crucial way to analyze user behavior habits 1. Since each person’s activity trajectory is very different, we can call its unique location information the user’s “fingerprint.” Therefore, we can analyze the user’s unique “fingerprint” analysis to dig out the personal habits and emotional tendencies, so as to achieve the purpose of depicting the user profiling. Further predicting the location of the mobile object at a certain moment in the future can provide important decision-making basis for urban planning and management, urban public safety, urban emergency command, etc., and can also provide technical support for location-based service applications such as personalized information recommendation and advertisement placement 2. With the increase of massive mobile data, the knowledge contained in the geographical location information from the moving trajectory data has gradually become a hot issue for scholars. Wang Liang 3 proposed a method of urban moving trajectory pattern mining based on multi-scale spatial partitioning and road network modeling. The proposed multi-scale elastic multi-scale is proposed for the existing rule meshing method which performs sequ
Data Loading...
