The Classification Algorithm Based On Unbalanced Data Set And Its Application In Communication Intelligent Operation

The development of mobile internet is changing the communication operation mode from network-oriented mode to customer-oriented mode.Since accurate marketing and customer chum warning for communication intelligent operation are typical cases of unbalanced data classification,the research of classification for imbalanced data classification and its application in communication intelligent operation is of great significance.This thesis focuses on the research of classification algorithm of unbalanced data set.Firstly,since SMOTE does not take data distribution into consideration when generating new samples and discarding the information of majority sample,an improved SMOTE algorithm based on neighbor sample distribution and Poisson distribution is proposed to solve the problem of unbalanced data set,which adopts a parameter related to the distribution of data to adjust the process of the interpolation.The simulation results verify the effectiveness of the proposed algorithm.Secondly,an improved kNN algorithm based on pre-classification is put forward to reduce the complexity of kNN algorithm.The data samples whose characteristics is not obvious are deleted in order to reduce the algorithm complexity,the simulation results verify the effectiveness of the algorithm.Thirdly,from the perspective of the communication intelligent operation,the proposed two algorithms are applied to construct a Model of Intelligent Operation for Communication based on Unbalanced Data set(MIOCUD)for the classification and prediction on customer churn data and marketing data from a communication operator.The simulation results reveals that the two algorithms provide an effective solution for communication intelligent operation.