Network Resource Allocation Strategy Based On Campus Mobile User Behavior Analysis

With the rapid development and popularization of the wireless Internet,the scale of coverage continues to expand,and wireless network applications are increasingly diversified,such as Web search,IPTV,video calls,online games and shopping,and live broadcasts,etc.As a result,the demand for network bandwidth resources is increasing.The export bandwidth of wireless network is limited,how to make efficient use of limited bandwidth while ensuring users' QoS requirements,that is,how to develop a reasonable wireless network allocation strategy,is of great significance.Focus on the research of wireless network resource allocation strategy in campus environment,the analysis and prediction of campus mobile user behavior is conducted,the law of users' mobility behavior and resources demand is studied.Based on the predicted resources demand for different businesses,the dynamic wireless resource bandwidth allocation strategy is proposed,which ensures users' QoS requirements with limited bandwidth and optimizes user experience.1)Based on the campus network of Southeast University,we introduce the deployment of AP,the interaction process between the user's mobile terminal and the AP,the campus network authentication mode,the campus network topology and related data collection methods.Then we verifies the feasibility of constructing the user's movement trajectory based on the AP access log information.2)After preprocessing the AP access log information,we construct the mobile model and studied the law of movement according to the user's movement trajectory in the campus.According to the similarity of the user's mobile behavior,the user is clustered by the fuzzy clustering algorithm to establish the state transition probability matrix of the user cluster.Based on the above research,we have proposed a user moving position prediction method based on fuzzy clustering.3)After preprocessing the campus wireless network data traffic data,we would propose the field information of HTTP protocol,and classify the user access behavior.Based on the time and location of user access behavior,we propose a user access behavior prediction algorithm based on multi-dimensional Markov model.We use the weight formula to weight the probability transfer matrix to improve the prediction accuracy of the model.4)Combining the two data sets of AP access log information and wireless network traffic data,we extract the characteristics of user mobile behavior and access behaviors.In this thesis,we propose a BPNN-based conditional reclassification algorithm to classify the campus network user identity,and verify the effectiveness of the algorithm from three aspects: accuracy,accuracy and recall.5)Based on the analysis and prediction of user's mobile behavior and access behavior,combined with user identity,we propose a dynamic bandwidth allocation algorithm based on user identity and behavior for the unreasonable problem of fixed bandwidth allocation in traditional networks.According to the QoS requirements of the user,the network bandwidth is dynamically adjusted to maximize the utilization of the wireless network bandwidth.The wireless network bandwidth utilization is considered on the basis of ensuring differentiated QoS requirements of different users.