论文总字数：32681字

摘 要

随着电商的飞速发展，逐渐开始出现倒逼线下的状况，巨大的压力让线下零售开始不断调整方向。与此同时，线上购物也面临着用户增速放缓，流量红利逐渐消失的问题。线下零售利用虚拟空间无法拥有良好的用户体验的劣势成为自己的优势，因此，线上线下结合才是“零售电商”的变革。“新零售”推动了线上线下和物流相结合，运用大数据、云计算等新技术为线上线下消费者提供全方位的消费服务。本文以“盒马鲜生”为新零售业态的代表，通过对“盒马鲜生”的UGC进行文本挖掘，构建一个可推广的客户满意度评价指标体系，并对“新零售”提出相应的建议。

首先，考虑到“盒马鲜生”APP的用户评论在本文研究期间的用户评论数量非常少，因此本文选择“大众点评”的“盒马鲜生”用户评论信息作为数据来源。 利用Python采集了好评39807条，差评3616条，并对评论进行分词、建立专业词典、同义词合并以及词频统计处理，并绘制词云图，分析“盒马鲜生”的消费者的关注重点。然后，为了使指标更为准确，采用基于层次分析法的德尔菲法，计算出各因素的权重，剔除权重排名靠后的指标，缩减指标数。最后，运用贝叶斯网络计算顾客满意度影响因素的重要程度，其中通过独立性检验计算因素之间是否存在依赖关系，得到基于该数据独立性约束的最优贝叶斯网络结构。计算各因素在评论中被提及的概率以及计算各因素在不同情感中出现的概率。这些计算结果可以清晰的展现出顾客的关注重点。

通过贝叶斯建模结果，发现“商品品质”、“商品种类”以及“售后效率”是盒马鲜生消费者最关注的因素。除此以外，结果还体现了因素之间的传导关系，比如品质影响是否需要售后。并对针对这些结果，提出相应的建议及对策。

关键词：UGC；文本挖掘；新零售；德尔菲法；贝叶斯网网络

Customer Satisfaction Analysis of New Off-line Retail Stores Based on UGC Text Mining: Taking He Ma Xian Sheng as an Example

Abstract

With the rapid development of e-commerce, there has been a situation of downward pressure offline, so that offline retail began to adjust its direction. At the same time, online shopping is also facing the slowdown of user growth and the gradual disappearance of traffic dividends. The problem that offline retailers can’t have a good user experience by using virtual space has become their own advantage. Therefore, the unification of offline and online is the change of "retail e-commerce". "New Retail" promotes the combination of online and offline and logistics, using new technologies such as big data and cloud computing to provide all-round consumer services for online and offline consumers. This paper takes "He Ma Xian Sheng" as the representative of the new retail format, through text mining of "He Ma Xian Sheng" consumer comments, calculates and analyses customer satisfaction with "Boxed Horse Fresh Life" offline stores, and puts forward corresponding suggestions for "New Retail".

Firstly, considering that the "He Ma Xian Sheng" APP does not open the user comment function, this paper chooses the "He Ma Xian Sheng" user comment information of "public comment" as the data source, and collects a total of 39807 comments, of which 3616 are bad. Using Python to crawl consumer reviews, and extract word frequency of reviews, statistics high-frequency vocabulary, analysis of "He Ma Xian Sheng" consumer focus. Then, in order to precise indicators, the Delphi method based on the analytic hierarchy process is used to calculate the weight of each factor, eliminate the indicators with the lowest weight ranking, and reduce the number of indicators. Finally, the Bayesian network is used to calculate the importance of the influencing factors of customer satisfaction, and the dependence relationship between the factors is tested by independence. The optimal Bayesian network structure based

on the data independence constraint is obtained. Calculate the probability of each factor mentioned in the comments and the probability of each factor appearing in different emotions. These results can clearly show the focus of customers' attention.

Through Bayesian modeling results, it is found that "commodity quality", "commodity type" and "after-sales efficiency" are the most concerned factors of fresh consumers. In addition, the result also reflects the transmission relationship between factors. If customers mention "commodity types" many times in a comment, then they must mention "commodity quality" many times. On the contrary, there will be fewer references to "store environment". And in view of the result of the play, the corresponding suggestions and countermeasures are put forward.

Keywords: UGC, text mining, new retail, Delphi method, Bayesian network

目 录

摘 要 I

Abstract II

第一章 绪论 1

1.1 研究对象 1

1.2 研究目的和意义 2

1.3 研究方法与框架 3

第二章 文献综述及关键技术 5

2.1 研究现状综述 5

2.1.1 “新零售”研究现状 5

2.1.2 基于UGC的文本挖掘研究现状 5

2.1.3顾客满意度研究现状 7

2.1.4 研究述评 7

2.2 相关关键技术介绍 8

2.2.1 网络爬虫 8

2.2.2 文本预处理 8

2.2.3贝叶斯网络概念 9

2.2.4贝叶斯网络结构学习 9

2.3 本章小结 10

第三章 数据处理及满意度指标的构建 12

3.1 数据来源及获取 12

3.2 数据处理及初步指标的提取 14

3.2.1 数据预处理 14

3.2.2 高频词统计及初步指标提取 15

3.2.3 数据转换为词频向量、汇总统计和离散化 18

3.3 基于专家调查法的指标确定 19

3.3.1 专家调查法实施 19

3.3.2 指标的确定 21

3.4 本章小结 22

第四章 基于贝叶斯网络建模的客户满意度分析 24

4.1 通过独立性检验计算各因素之间的依赖关系 24

4.2 各因素与评价之间关系 28

4.2.1 计算各因素与评价的条件概率 28

4.2.3计算各因素与评价的后验概率 29

4.3 基于德尔菲法的贝叶斯网络建模计算结果的对策及建议 31

4.4 本章小结 32

第五章 总结与展望 33

5.1 研究总结 33

5.2 不足及展望 33

参考文献 34

致 谢 37

附 录 38

请支付后下载全文，论文总字数：32681字