The Research And Application Of The Big Data Mining In Communication Network Prediction

Mobile communication is one of important tools for national life,which is needed to be stability and security anytime. Network optimization is a vital method to complete this task and network forecasting is the precondition of network optimization. According to the traditional network prediction methods lack of user data and based on assumption network scenario, this paper suggests to apply the big data technology to network key performance index(KPI) forecast,which is in the form of time series. At the same time, support vector machine(SVM) and association rules technology are used to optimize forecast process and results in order to faster rate and higher accuracy.Firstly,this paper summarizes the various communication network prediction approaches status,significant and thought, which has laid the groundwork for follow-up study.Secondly, it briefly expounds the mobile communication network and it’s related knowledge,such as KPI. Then on this basic it introduces the related concepts of time series, the advantages and disadvantages of common prediction algorithms and application of the big data in time series.Thirdly, a new time series prediction algorithm based on the big data is put forward according to the characteristics of network performance and the defects of traditional methods.The algorithm imports the concept of sudden components,which is based on traditional three time series feature components extraction.Also, the paper detects abnormal data and makes predicted analysis according to the result of extraction.Fourthly, in order to solve low rate problem when deal with huge amounts of data and further improve the precision, the article has discussed how to apply the support vector machine and association rules to the big data prediction algorithm.The paper divides initial data into busy and no-busy by SVM, and predicts different data segment respectively in order to improve rate. At the same time association rules is used to determine correlation function for optimizing result.Finally, the paper takes the Nanjing area real network KPI to forecast analysis according to algorithm. The experimental results show that the modified big data prediction algorithm has higher precision and rate when compared with traditional time series prediction algorithm.