基于评论数据的B2C客户消费偏好模型研究

发布时间：2018-05-06 20:13

本文选题：大数据 + B2C　；参考：《安徽理工大学》2017年硕士论文

【摘要】：互联网的迅速发展使得网络购物消费快速增长,近年来以京东、天猫等为代表的B2C购物模式发展迅速,网站业务量和信息量迅速增加给企业发展带来挑战。如何从逐渐增加的非结构化数据中提炼有效信息?如何从海量消费数据挖掘客户的真实需求从而提供精准的个性化服务,最大程度改进客户的购物体验?这些问题成为目前研究的热点和难点。因此,运用数据驱动模式挖掘客户的消费偏好,是B2C购物网站精准营销的重要保障。本文以在线评论、消费者行为和B2C网站客户消费偏好为理论基础,以天猫B2C服装类客户消费作为研究对象,从消费者、平台及商家方面分析消费偏好影响因素,对所选定网售商品进行归类和筛选,确定了 7种服装商品,运用爬虫软件抓取2016年9-11月的在线评论信息。通过数据整理、关键词提取与统计分析等手段,提取客户评论信息的34个高频关注点,确定12个特征因素变量。运用李克特量表的5级评分标准将评论信息转化为结构化数据。运用Clementine12.0软件将12个商品特征因素变量导入,建立各个因素之间的贝叶斯网络模型结构。计算各节点在其父节点条件下的条件概率分布,各特征因素重要度,建立logistic回归模型,对比分析贝叶斯网络模型的准确性,对模型预测结果做出准确评估。结果表明,所筛选7个商品的舒适程度、面料、质量、颜色、合适程度、价格等,都是客户高频关注词;贝叶斯网络模型中因素节点间具有较强的相关性;节点的条件概率分布情况相似,客户给予优、良、中评价的概率较高;男装和女装的特征因素重要性程度不同,女装较关注物流、相符程度、手感、正品、合适程度等因素,男装则关注面料做工、色彩、物流、手感、美观程度、舒适程度等因素。B2C网站可根据客户消费关注高频词,贝叶斯网络因素关联,各种因素所得评价分数的概率以及重要度分析消费偏好,制定精准营销策略。
[Abstract]:With the rapid development of Internet, the consumption of online shopping increases rapidly. In recent years, the B2C shopping model represented by JingDong and Tmall has developed rapidly, and the volume of business and information on the website is increasing rapidly, which brings challenges to the development of enterprises. How do you extract valid information from the increasing amount of unstructured data? How to mine the real needs of customers from mass consumption data to provide accurate personalized services to maximize the improvement of customer shopping experience? These problems have become the hot and difficult point of current research. Therefore, it is an important guarantee for accurate marketing of B2C shopping website to use data driven mode to mine customer's consumption preference. Based on online review, consumer behavior and consumer preference of B2C website, this paper takes Tmall B2C clothing consumer as the research object, analyzes the influencing factors of consumer preference from consumers, platforms and merchants. This paper classifies and selects the selected online items, determines 7 kinds of clothing products, and uses crawler software to capture the online comment information of September-November 2016. By means of data collation, keyword extraction and statistical analysis, 34 high frequency concerns of customer comment information were extracted, and 12 feature factor variables were determined. The comment information was converted into structured data using the 5-level rating scale of the Richter scale. Using Clementine12.0 software, 12 commodity feature variables are imported and the Bayesian network model structure between each factor is established. The conditional probability distribution of each node under the condition of its parent node and the importance of each characteristic factor are calculated. The logistic regression model is established and the accuracy of the Bayesian network model is compared and the prediction results of the model are evaluated accurately. The results show that the comfort, fabric, quality, color, suitability and price of the seven products are the high-frequency concern words of the customer, and there is a strong correlation among the factors nodes in the Bayesian network model. The conditional probability distribution of the node is similar, the probability of customer giving excellent, good and medium evaluation is higher; the importance of characteristic factors of men's wear and women's wear is different, and women's wear is more concerned with the factors such as logistics, match degree, hand feeling, genuine product, suitable degree and so on. Men's clothing is concerned about fabric workmanship, color, logistics, feel, beauty, comfort and other factors. B2C website can be based on customer consumption concerns about high-frequency words, Bayesian network factors related, The probability and importance of evaluation scores are analyzed and precise marketing strategies are formulated.
【学位授予单位】：安徽理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：F724.6;F713.55

【参考文献】