Research On Mining Traveling Partners Of Moving Objects And Location Prediction

With the rapid development of the technology of the Internet of Things and moving objects database technology in recent years, a large number of moving objects data is collected by various types of sensor equipments. Research results in the field of moving objects have been used in many areas, such as checking of moving social applications, transportation and logistics management, biological migration research, the detection of military intrusion and so on. Many major companies develop many kinds of practical applications in moving environment. For example, these applications will recommend some goods and hotels based on user’s location, and it also gives you the best information by considering user’s needs and characteristics. Because sensor devices generate data on TB level in every day, the main problem of moving objects study is how to deal with so much massive moving objects data. Recently, researchers begin to care about mining frequent path model in large number of moving objects data, discovering moving objects that move together is one of hot issues.Traditional methods are difficult to efficiently discovery the model of moving objects group that move together. Either these methods ignore the information of time dimension, or they have excessive consumption in space and time, so they don’t use in the actual scenario. On basis of studying existing methods at home and abroad, this paper mainly carried on deep research in extracting time feature points from moving objects trajectories, discovering traveling partners and predicting the future location of moving objects. The main contents and innovations of this dissertation are as follows:1) This paper introduced the concept of information entropy and information gain aimed at the problem of massive and uncertain moving objects data. By calculating information gain of each time point, time feature points are extracted from nodes with the maximum information gain. The method can greatly reduce spatial-temporal consumption. It can report key nodes that are able to represent the movement trend to the system by providing the basis of selecting time sample points for discovering traveling partners at the next step.2) In order to discovery the model of traveling partners from RFID data that are generated by moving objects, this paper introduced the clustering and intersecting algorithm (CI) and proposed a closed clustering and intersecting algorithm (CCI) to discovery traveling partners that move together. The algorithm is mainly constituted by two steps:first step is clustering sub-trajectory, it generates sub-trajectory clusters; second step is intersecting sub-trajectories with the traveling partners’candidate set to improve the candidate set, and find out traveling partners. In this process, we use the principle of Closure to reduce the time consuming and accelerate our processing in discovering traveling partners. Experiments show that the algorithm can quickly and effectively find relevant traveling partners, so it is suitable for solving the issue. 3) We propose a location prediction method based on spatial-temporal suffix tree model aimed at motion characteristics of moving objects. Firstly, we analyze historical trajectory data to find out relevant frequent area. And then convert to these trajectories in order to establish a spatial-temporal suffix tree model (STS-tree). Finally, we use the recent trajectory sequence information and query time to predict next location information of moving objects at a certain time in future. This result can help us to judge the future movement trend of moving objects for analyzing and establishing some decisions. Experiments demonstrate that the clustering method based on grid is more reasonable and efficient than one based on density in the problem of discovering frequent areas. In the location prediction, STS-tree prediction method is also better than the conventional hybrid prediction method.