社会化商务中的商家信誉分析及评价研究

发布时间：2018-05-01 04:41

本文选题：社会化商务 + 信誉评价　；参考：《大连理工大学》2016年硕士论文

【摘要】：社会化媒体与电子商务相互渗透、广泛融合,催生了兼具二者双重属性的社会化商务。社会化商务背景下,商家数量的快速增长与商家信誉水平的参差不齐,使得商家信誉评价问题越发紧迫与重要。而用户评论文本是蕴含商家信誉信息的宝藏,通过萃取和语义分析等文本挖掘技术能够快速获得其中的有效信息,为消费者做出购买决策提供依据,为商家改善服务、提高信誉水平提供参考。针对上述问题,本文通过对评论文本抽取主题词,采用聚类分析以及情感分析等文本挖掘方法,构建商家信誉维度体系,对商家信誉水平进行评价,主要内容如下：商家信誉维度体系构建。提出评论文本主题词抽取方法,将主题词抽取分为初步抽取、主题词扩展、词频调整三个步骤,以保证主题词抽取完全；提出基于HNC的主题词聚类算法,将HNC语义信息引入改进的CURE算法,获得更好的聚类效果；以代表点描述聚类簇作为信誉维度,根据聚类簇中的主题词词频计算各维度权重,生成信誉维度体系。最后实验验证了本文提出的评论文本主题词抽取方法以及主题词聚类方法的准确性,并以京东平台手机评论文本为实例,构建商家信誉维度体系,证明了方法的可行性与有效性。商家信誉评价。将信誉维度体系中的主题词还原到评论文本,通过依存句法定位修饰主题词的情感词、程度副词和否定词；提出基于HNC的情感词的情感值计算方法,基于距离的否定词权重计算方法,并建立程度副词权重表,确定主题词的评价情感值；结合词频信息计算各信誉维度的信誉值,结合各维度权重值计算商家的综合信誉值。最后实验验证了本文提出的情感值计算方法与人工标注的吻合度,在实例分析中采用本文提出的方法计算商家的信誉值,并对商家的信誉水平进行了分析和评价。本文提出的商家信誉维度体系构建方法以及信誉评价方法,丰富了文本挖掘理论的研究,弥补了目前社会化商务背景下商家信誉评价工作的不足,具有良好的应用前景。
[Abstract]:Social media and e-commerce are infiltrated and integrated widely, which have brought about a social business with the dual attributes of two parties. Under the background of socialized business, the rapid growth of the number of businesses and the uneven reputation of the merchants have made the reputation evaluation of merchants more urgent and important. The text mining technology, such as extraction and semantic analysis, can get the effective information quickly, provide the basis for the consumer to make the purchase decision, and provide the reference for the business to improve the service and improve the prestige level. This method constructs the business reputation dimension system and evaluates the reputation level of the merchant. The main contents are as follows: the business reputation dimension system is constructed. The extraction method of comment text theme words is put forward, and the extraction of subject words is divided into three steps: the initial extraction, the extension of the subject words and the word frequency adjustment to ensure the complete extraction of the subject words; based on the HNC The HNC semantic information is introduced into the improved CURE algorithm to obtain better clustering results. The representative points are used to describe the clustering cluster as the reputation dimension and calculate the weight of each dimension according to the word frequency of the topic word in the cluster. Finally, the method of extracting the topic word extraction method proposed in this paper is verified. And the accuracy of the topic word clustering method, and taking the Jingdong platform mobile comment text as an example to construct the business reputation dimension system, which proves the feasibility and effectiveness of the method. The emotion value calculation method based on HNC, the calculation method of emotional value based on the emotion words, the weight calculation method based on the distance based negative words, and the establishment of the degree adverb weight table, to determine the evaluation emotion value of the subject word, calculate the reputation value of each credit dimension with the word frequency information, and calculate the comprehensive reputation value of the merchant with the weight value of each dimension. Finally the experiment is tested. In this paper, the agreement between the calculation method of emotion value and the manual annotation is proved. In the case analysis, the reputation value of the merchant is calculated by the method proposed in this paper, and the reputation level of the merchant is analyzed and evaluated. The method of building the reputation dimension system and the reputation evaluation method proposed in this paper enriches the theory of text mining. The research makes up for the shortage of business reputation evaluation under the background of social commerce and has good application prospects.

【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP391.1

【相似文献】