Abstract:This paper mainly combines the demand of Zhejiang Huzhou power business, aiming at breaking the blind spot where customers have complaints about power consumption, thereby improving the management level of users'' electricity demand and realizing the reasons mining for hot complaints business work orders. In order to better dig deeper into the information contained in the complaints worksheet, based on natural language processing technology, this paper makes a in-depth text mining of power customer complaints work orders, uses hidden Markov model and other word segmentation techniques to analyze the acceptance of complaints in the work order content, and carries out the word frequency statistics, the keyword importance weight value is calculated by TF-IDF algorithm, and the keyword frequency with large weight value is extracted as the final result of customer complaint text mining. Finally, the word cloud analysis technology is used to visualize the analysis results, Through the text classification analysis, the text classifier model is constructed to realize the research on the distribution of "hot word frequency" in different business, and the corresponding improvement measures are carried out according to the results.The main problems of current power customer complaints are controlled, and differentiated service strategies are provided for different types of power customers, thereby improving customer satisfaction and loyalty. The promotion and application of the topic can improve the work efficiency of the customer service department, and solve the problem for the customer.