Mobile Communicaiton Network Optimization And Operation Technology Based On Data Mining Algorithms

With the rapid development of the mobile Internet and the extensive permeation of smart devices, great impact has been brought to the way of network quality assurance and business operaton for mobile communication operators. How to evaluate to which degree radio resource configuration scheme and business demand math with each other? How to dynamically allocate the definite radio resource to improve the network capacity? How to deeply dig the subscribers’behavior pattern to provide better experience? These issues are urgent to tackle.This thesis is focused on two key problems, i.e., the analysis and optimization of radio resource utilization efficiency, and the analysis and recommendation of subscribers’application preference. By way of deeply mining the valuable data generated in the network operation process, solutions to radio resource allocation optimization from the network’s aspect and intelligent application recommendation from the subscriber’s aspect are proposed, thus improving the network service capacity, network quality and subscriber’s experience. This thesis is divided into two parts. In the first part, the current situation of radio resource utilization is analyzed and an optimization scheme for dynamic radio resource allocation is proposed to improve resource utilization efficiency. First, data about network resource configuration and traffic load are collected from the operator’s network management system. Then, the temporal and spatial characteristics of spare radio resource are analyzed by means of cell clustering, classification, correlation and other data mining techniques. Analysis results show that radio resource and traffic demand do not match with each other both temporally asn spatially, proving the room for improvement. Naturally, an optimization scheme for dynamic resource allocation based on traffic demand prediction is put forward. In this scheme, gaussian weighted k-nearest neighbors algorithm is designed to predict the traffic demand and an optimization algorithm is proposed to allocate radio resource dynamically, thus improving the resource utilization efficiency and balancing the network traffic load. In the second part, subscriber’s behavior, especially application preference is analyzed and intelligent recommendation algorithm is designed to figure out subscribers’most favorite applications, thus laying the foundation for elaborative marketing strategies. First, deep packet inspection technique is adopted to collect data about subscribers’ application preference. Then, subscribers’profiles are figured out from the aspects of frequency of usage, traffic volume and so on. At last, singular value decomposition, clustering and nearest neighbor algorithms are combined together to generate an application recommendation system in order to figure out the most favorite application types for subscribers.Finally, the thesis is concluded by pointing out the deficiencies of the the proposed scheme and indicating the direction for improvement in future work.