Research On Customer Complaint Prediction Model Based On State Grid Service

With the deepening of electric power reform and the gradual liberalization of the electricity sales market,it is necessary for electric power enterprises to take market-oriented and customer demand as the center to speed up the overall improvement of power supply service and the promotion of marketing service level.According to the unreasonable arrangement of customer service personnel and serious complaints in the State Grid Electric Power Company,this paper proposes to build a customer complaints prediction model based on the big data of the State Grid Service,so as to realize the prediction of customer complaints in the next week.The research work in this paper is as follows:1.Three representative data mining algorithms,ARIMA time series,multiple linear regression and BP neural network,are selected to construct customer complaint prediction models respectively:ARIMA(p,d,q)model obtains stationary series by d-order difference of time series,determines the value of autoregressive term P and moving average term Q according to autocorrelation and partial autocorrelation function,filters 25 initial independent variables by stepwise regression method,obtains 10 independent variables,estimates model coefficients by least square method,and obtains multivariate linear regression.Regression equation;BP neural network model is based on PCA dimensionality reduction to get ten principal components,the number of output layer nodes is 10,and then design the network topology structure,for a hundred network training.By comparing the relative errors between the predicted values and the real values of the three models,it is concluded that the relative errors between the predicted values and the real values of the BP neural network model are 95.26% and less than 40%.2.Aiming at the problem that BP neural network is easy to fall into local minimum value,this paper introduces genetic algorithm,through which the original population is coded,initialized,selected,crossed and mutated,and the optimal initial weights and thresholds of BP neural network are obtained.Then the BP neural network is trained,and a customer complaint prediction model based on GA-BP neural network is constructed.Comparing the relative errors between the predicted value and the real value of the network and the GA-BP neural network,it is concluded that 99.12% of the predicted value and the real value of the GA-BP neural network have relative errors less than 40%.Therefore,GA-BP neural network model is the best customer complaint prediction model.The model is applied to customer complaint prediction,which effectively solves the unreasonable arrangement of customer service personnel and the serious problem of complaint.