Prediction of Business User Segmentation Model Based on Customer Value

With the continuous deepening of China's power system reform, the diversification of power customer demand is becoming more and more obvious. Carrying out customer segmentation is conducive to providing better service to customers by power companies and achieving a win-win situation for both enterprises and customers. Based on the customer segments in the power industry and the proprietary features of commercial users, this paper uses the unsupervised learning algorithm (Expectation-Maximization algorithm) in big data mining technology to build a business user behavior segmentation model to solve the problem of poor operability of combining qualitative and quantitative indicators and proposed a new theory system of subdivision for business users. Aiming at the shortcomings that the EM algorithm is sensitive to the initial parameters, the EM algorithm is proposed to improve and optimize the model and enhance the accuracy of customer segmentation.