Research On Prediction And Optimization Technologies Of User Behavior In Wireless Ubiquitous Environment

In wireless ubiquitous environment, the user-centric and application-driven network evolution trend has become inevitable. It is necessary to predict user service behaviors and then recommend the most suitable network resources, for guaranteeing and optimizing QoE of each user. Therefore, prediction and optimization technologies of user behavior appear and attract extensive attention, which will be investigated in this thesis. The main works are as follows:I. The related contents of wireless ubiquitous environment are introduced, and then current prediction and optimization technologies of user behavior are surveyed respectively. Furthermore, a user behavior prediction and optimization system is designed, and then the corresponding function module descriptions as well as actual development considerations are given.II. A variable order Markov model based user behavior prediction algorithm is proposed. Firstly, a user similarity based fuzzy clustering algorithm is presented to categorize users. Then, a variable order Markov model based on probabilistic suffix tree(PST) is introduced to predict the future service states of users. And a fusion prediction model is established to fuse prediction results, so as to improve the prediction accuracy. Simulation results show the effectiveness of proposed algorithm.III. A weighted Markov model based user behavior prediction algorithm is proposed. Firstly, an optimal order selection algorithm based on information entropy is presented for determining the optimal order of Markov prediction model. Furthermore, by using an adaptive feedback based weight correction method, a modified weighted Markov model is formulated to accurately predict user future service states. Simulation results show the effectiveness of proposed algorithm.IV. A multi-terminal scenario oriented user behavior optimization algorithm is proposed. Firstly, based on cooperative technology, a multi-terminal service information aggregation mechanism is presented to build the complete historical information trie of user service states, for accurately predicting user future service states. Secondly, according to the prediction results, the corresponding heavy-tailed model is used to estimate the service duration. Moreover, by integrating service preferences and terminal attributes, a matching metric of services and terminals is established, and then a terminal recommendation model is proposed for recommending the optimal service terminal to each user. Simulation results show the effectiveness of proposed algorithm.Finally, the conclusions are summarized, and the future research directions are given.