投诉信息自动分类与推送系统的研究与设计

发布时间：2018-12-20 10:32

【摘要】：随着通信服务发展的深入,出现了许多新问题和新矛盾,这些问题最终都会转化为客户投诉到达投诉分析部门。能够有效地分析投诉,快速的抽取出用户、分析部门或其他业务部门关注的投诉信息,无疑对正确的处理投诉、快速的发现问题十分有帮助。然而目前还没有令人十分满意的以用户关注为中心的投诉信息识别技术,多数情况下为了保证准确,仍然只能依靠人工识别,不仅花费大量的人力成本和时间成本,而且也无法较好的适应数据量大、数据量增长速度快的情况,容易造成数据的积压。更重要的是,容易错失及早发现问题的良机,使投诉信息反应的问题无法在进一步扩大之前得到遏制。这样不仅会降低客户的满意度,严重时甚至会使客户对品牌形象丧失信心。针对以上问题,本课题通过对投诉数据进行分析,并对投诉分析人员筛选目标投诉的行为进行研究和记录,力图使用技术手段解决以投诉分析人员的关注为核心的投诉信息自动分类和推送的问题。同时结合实际的项目需求,将其应用到投诉信息自动分类与推送系统中,为投诉分析人员提供一个筛选目标数据的辅助平台。本文主要完成以下工作：一、通过对投诉数据和用户分析习惯的研究,明确本文所提的自动分类和推送的定义和相关概念。二、提出了一种基于TF/IDF算法的文本特征抽取方法,并针对正向文本的文本特征,利用关联规则算法挖掘类特征词集。三、提出了一种根据VSM模型和全部类别类特征词集构筑投诉空间的方法。在此基础上,提出了由投诉空间、分类算法和分类参数共同组成的分类模型。四、通过记录用户标注样本的行为,抽取推送关系映射,并在此基础上构建自动推送模型,以达到正确推送分类结果到对应用户的目的。
[Abstract]:With the development of communication service, there are many new problems and contradictions, which will eventually turn into customer complaints to the complaint analysis department. Can effectively analyze complaints, quickly extract users, analysis departments or other business departments concerned about complaints information, no doubt for the correct handling of complaints, the rapid discovery of problems is very helpful. However, at present, there is no very satisfactory complaint information identification technology centered on the attention of users. In most cases, in order to ensure accuracy, we still have to rely on manual identification, which not only costs a lot of manpower and time, but also costs a lot of time. And also can not adapt to the large amount of data, the rapid growth rate of data, easy to cause the backlog of data. More importantly, it is easy to miss opportunities for early detection of problems, so that the problem of response to complaints information cannot be curbed before it can be further expanded. This will not only reduce customer satisfaction, and even make customers lose confidence in brand image. In view of the above problems, this subject analyzes the complaint data, and researches and records the behavior of the complaint analysts to screen the target complaints. This paper tries to solve the problem of automatic classification and push of complaint information with the attention of complaint analysts. At the same time, it is applied to the automatic classification and push system of complaint information according to the actual project requirements, which provides an auxiliary platform to screen the target data for the complaint analysts. The main work of this paper is as follows: first, through the research of complaint data and user analysis habits, the definition and related concepts of automatic classification and push are clarified. Secondly, a text feature extraction method based on TF/IDF algorithm is proposed. Thirdly, a method of constructing complaint space based on VSM model and all kinds of feature word sets is proposed. On this basis, a classification model composed of complaint space, classification algorithm and classification parameters is proposed. Fourthly, by recording the behavior of the user tagging samples, extracting the push relation mapping, and constructing the automatic push model on this basis, to achieve the purpose of correctly pushing the classification results to the corresponding users.
【学位授予单位】：北京邮电大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TP391.1

【参考文献】