Research On Mining The User Mobility Patterns And Location Prediction

With the continuous development and extensive applications of technologies in the fields of both communication and computer network, mobile computers can access kinds of network at any time and any place to get the data information they needed in the near future. This special distributed computing environment is called mobile computing. The mobile computing is considered as one of the most influential technology in the future and has become the hotspot of research.In mobile computing environment, using predicted location information of a mobile user to design rational and effective mobility management schemas is one of the challenging topics. How to effectively predict a mobile user's location is the key to achieve mobility management that based on the predicted location. On the other hand, mobility decisive problem is important to a mobile user in mobile computing environment. For a mobile user, how to select the optimal paths to reach the destination in terms of the network QoS and road condition is an important part of mobile decisive problem.Regarded the national natural science program, the research ofmobile decisive theory and its support system, as its research background, this dissertation proposed two different methods used to predict the locations of a mobile user. If we have few location information of a mobile user, we proposed the local linear forecast method used to predict the moving trajectories in real time. Simulations of a mobile user's movement in one dimension and two dimensions are done. If a great deal of the location information is exists, we proposed a non-real time mobility prediction algorithm based on data mining. In terms of many moving logs stored in the visited location registers of the network, we can mine the mobility patterns, which will be used to predict a mobile user's location. The whole algorithm consists of three phases. In the first phase, based on the former Apriori algorithm, we proposed Apriori incremental algorithm to deal with the increasing database. For the low efficiency of Apriori algorithm, PrefixSpan algorithm is proposed to mine mobility patterns. Compared with Apriori algorithm through examples, we draw the conclusion that two algorithms have the same results, but PrefixSpan mining algorithm is more efficient. In the second phase, the mobility rules are extracted from the mobility patterns. In the last phase, the mobility rules, which match the current trajectory of a mobile user, are used for the prediction of a mobile user's next movement. At the same time, we discussed a mobile user's moving problem in a concrete spatial environment. We did some researches in the optimal paths selection of amobile user by considering both the network QoS and the road lengths with a fixed destination. We constructed the moving models, proposed the method of giving weight to the QoS and road length and comparing the path weight that start with the current point to select the optimal paths.It should be pointed out that theories, methods and results introduced in this dissertation are quite useful in such research areas as location updating and paging based on predicted location information in mobility management, which aims to improve mobility management performance, reduce signal instructions and increase the effective capacity of the system, the improvement of current mobile communication networks, and the design of the next generation mobile communication networks. At the same time, with a fixed destination, by selecting optimal paths, a mobile user can practice quick response and quick transaction to the bursting out events, which has an important meaning to the emergency decisive system such as military, disaster transaction and so on.