Application Of Data Mining Algorithm In The Network Optimization Of Traffic And Poor Cell Mining

Data mining is an application technology which can process huge amounts of data. It is very popular in many fields and it also has many functions,such as correlation analysis,prediction,classification,clustering and so on. It extracts useful information by finding out the potential relationship of data. This thesis selects correlation analysis and hierarchical clustering method to study the application in mobile network optimization. Then it puts forward the traffic prediction model and uses clustering method to analysis the poor area, and also gives the idea of optimization.The thesis first introduces the process of network optimization and related technology of data mining. At the same time, it analyzes whether the data mining technique can use in the network optimization work. It analyzes the busy traffic prediction in holidays firstly. Considering there are many factors affect the prediction accuracy of traffic, so it selects the Multi-factors Grey traffic based on Support Vector Machine forecasting model to predict the short-term changes in traffic. The author uses the Multi-variable Grey model to forecast the basic rules of traffic, and then uses the Support Vector Machine model to amend the prediction error. In order to further improve the prediction accuracy, the author then puts forward an improvement scheme, a complex Gaussian wavelet kernel function which has strong convergence and divergence and global optimization ability to optimize the vector machine. The simulation results show that the model has better convergence effect and prediction results. Finally, this article puts forward a hierarchical clustering method on cell performance division. Use sample cluster analysis method to find the worst performance cell, and then analyze the correlativity between the poor performance cell and the performance indicators by using variable cluster analysis method. People can formulate optimization scheme according to the result.