Research And Design Of Telecommunication Customer Churn Prediction Algorithm Based BP Network

Telecom market environment will be increasingly fierce competition at home and abroad.The increasing of telecommunications and frequent telecom fraud,are forcing the telecommunications industry to adopt effective strategies to the survival and development.Although the level of services of the telecom industry has obvious increase,but in the face of competition increased,in the face of limited marketing resources and mobile users demand changing,the customer churn is inevitable.Therefore,the customer churn prediction is a major concern of telecommunications.Based on the Liaoning CMCC datasets,this paper puts forward two kinds of valid customer churn prediction algorithm.Owing to the datasets having too many attributes,the traditional prediction method based on statistics is not applicable,because the BP neural network can approximate any complex nonlinear function,so this paper proposes a customer churn prediction model based on BP neural network.According to the characteristics of dataset,this paper establishes the structure and parameter,and training function of network.Customer churn prediction model based BP neural network has achieved well enough results.Due to the BP neural network uses gradient descent for training,the local search ability is strong,but the global search ability is relatively weak.According to this problem,this paper puts forward with the capability of global search of improved particle swarm optimization algorithm and improved genetic algorithm to optimize the BP neural network.In this paper,an improved PSO based BP network for telecommunication customer churn prediction algorithm is proposed.The improved PSO algorithm classifies the particles into three categories according their fitness value,and updates the velocity of different category particles using distinct equations.And take the BP network as the fitness function of the particles.The improved PSO initials the weights and thresholds of the BP network,and brings remarkable improvement on customer churn prediction accuracy.Similar to the improved PSO-BP algorithm,in this paper,an improved GA based BP network for telecommunication customer churn prediction algorithm is proposed.The improved algorithm proposes a self-adapting probability of crossover and mutation,enhance the global optimum search ability of GA,and the improved GA-BP has obvious improvement on customer churn prediction compared with existing algorithms.The improved GA-BP has shorter training time compared with improved PSO-BP.