基于自然语言的国网投诉工单智能分类模型构建  

Construction of Intelligent Classification Model of Complaint SheetstoState Grid Based on Natural Language

作　　者：张兆芝 陈翔 高敏 卢燕燊 张钟杰 ZHANG Zhaozhi;CHEN Xiang;GAO Min;LU Yanyan;ZHANG Zhongjie(Fujian Power Supply Serice Co.,Ltd.,Fuzhou,Fujian 350000,China)

机构地区：[1]福建省供电服务有限责任公司,福建福州350000

出　　处：《微型电脑应用》2020年第7期54-57,共4页Microcomputer Applications

摘　　要：为了更好的深入挖掘投诉工单背后所蕴含的信息,从自然语言处理技术出发,对客户投诉工单进行深入文本挖掘。在对电力投诉工单进行数据清洗的基础上,运用jieba进行分词,构造专业词典提升分词准确度,并对特征进行降维,然后运用利用词袋模型对中文文本进行分词,利用Bagging集成模型,构造包括朴素贝叶斯模型、决策树模型等在内的多个分类器模型,实现对词频在不同业务中的分布情况的研究,并根据结果开展相应改进措施,把控住当下电力客户投诉的主要问题,为不同类型的电力客户提供差异化的服务策略,以落在实处的为客户解决难题。In order to better dig out the information behind the complaint sheet,based on natural language processing technology,customer complaint sheet is deeply mined.On the basis of data cleaning for power complaint worksheet,the key dictionary is constructed,and the dimension of the feature is reduced.Then the Chinese text is segmented by using the word bag model,and several classifier models including Naive Bayesian model and decision tree model are constructed by using Bagging integrated model.The distribution of word frequencies in different services is studied.Result corresponding improvement measures are carried out to control the main problems of current power customers’complaints,and to provide different service strategies for different types of power customers,so as to solve the problems for actual customers.

关 键 词：自然语言 投诉工单 分类器模型 

分 类 号：TG409[金属学及工艺—焊接]