【作者】 田野；

【导师】 苏莉蔚；

【作者基本信息】 吉林大学 ， 软件工程， 2014， 硕士

【摘要】 在电商购物的用户的购买行为受到多种因素的影响，包括有内部和外部的。内部因素也就是用户的自身条件或对商品、购买行为的反映，这属于影响购买行为的主要问题；外部因素受限于电商为用户营造的购买环境，在实际调查中发现同样具有一定的影响。过去因为技术、方法、行业特性所限，互联网通过线上所吸纳的海量数据的确给电商企业带来了短期的处理空间，而直接从线下模式获取数据来源的成本及模式尚不成熟，但是市场最大的数据源永远都是在线下，电子商务的本质是通过电子化手段来服务传统商务流程，帮助其降低成本提升效率。所以直接建立起线下数据的采集接口渠道，而不再仅仅依靠纯线上数据作为来源是一个关键环节。同时，线上线下数据采集汇聚到数据中心，产生的分析型数据将会反哺线上，带来更精准的使用价值。本文通过E－mail、QQ留言、微信群方式将网络购物市场调查问卷发送给被调查者，被调查者完成后将结果通过对应方式返回。同时也在线下进行了一些面对面的问卷调查。共收回调查问卷结果近200份，其中多数为家人、同学、同事、朋友、网络好友等。本文研究的目的是利用数据挖掘算法和多种挖掘模式，针对所能够收集的样本数据，深入分析网购用户的行为，从中得出一定的规律，为电商提供参考意见，以辅助其市场定位。当然，本文属于试验性研究，也希望通过这样一次体验，学习和了解数据挖掘的技术，为今后更好的应用打下基础。具体研究内容如下：1.多种数据挖掘模式分析，包括聚类分析、分类/预测、关联规则。2.问卷调查设计方法。3.顾客行为分析。4.多种数据挖掘算法：TwoStep算法、K-Means算法、QUEST算法、Apriori算法。5.数据统计分析软件：SASS、Clementine数据挖掘平台。经过数据分析得到五个结论：1.半年内网购次数较多并且每次购物金额为100-500之间的顾客受网店信誉度的影响很大，但该类用户所占比例不高。半年购物次数较多、每次购物金额在100-500之间的用户受网店信誉度影响不是很大，并且这两类用户在群体中所占比重较大。2.半年内网购次数较多并且网购金额在100-500之间的顾客比较容易受网店搜索排名的影响，支持度较低表明该类顾客所占比例较低。表明多数顾客不受网店搜索排名的影响。3.网购历史为5年或5年以上的，半年内网购频率较高；网购历史在一年以内的较少网购；网购历史在1-5年的人群有三分之二的人偶尔网购，三分之一的人经常网购。4.网购的金额与网购历史为正比例关系，网购历史较长的人群消费金额为上涨趋势，并且较高；网购历史较短的人群消费金额趋于保守；网购历史为1-5年的人群较多，消费也趋于中等。5.网购潜在客户的人群大致可以分为以下几类：(1)登录互联网时间在四年以上的人群。(2)年龄在18-35之间的网民。(3)具有大专或本科学历的网民。在本文中还收集了权威互联网调查和管理机构发布的分析数据，用来验证本文的分析结论，结果证明基本准确。更多还原

【Abstract】 Customer purchasing behavior is influenced by external factors, namely externalstimulation; there are internal factors, namely personal characteristics. External factor is theexternal cause, for the customer to make a buying decision provides conditions, and theinternal factor is the basis of customer respond. The main factors influencing the buyingbehavior are the internal cause, that is, the customer’s personal characteristics.Technology because in the past, and the limit of the method, industry characteristics,absorb the huge amounts of data by the Internet through online indeed brought electricityenterprise short-term deal with space, and the cost of extracting data source directly fromoffline mode and pattern is not yet mature, but the market’s largest data sources is always online, the nature of e-commerce through electronic means to serve the traditional businessprocess, help them reduce cost increase efficiency. So established offline channels of dataacquisition interface directly, instead of just rely on pure online data as a source is a key link.At the same time, online data collection to data center, the analytical data will be feedbackonline, with the use of more accurate value.In this article, through E-mail, QQ messages, micro letter group way to networkshopping market survey questionnaire sent to respondents, after the completion of therespondents would return the results through the corresponding manner. And at the same timeoffline some face-to-face questionnaire survey was done. Total recovery of the questionnaireresults from nearly200, mostly family, classmates, friends, colleagues, friends, network, etc.The purpose of this study is to use data mining algorithm and a variety of mining mode,aiming at to collect the sample data, in-depth analysis of online user behavior, draw certainregularity, provide reference opinions for electricity, to assist its market positioning. Ofcourse, this article belongs to experimental study, also hope that through such an experience,to learn and understand the technology of data mining, to lay the foundation for theapplication of the future better. The concrete research content is as follows:1, Data Mining Patterns: Cluster analysis, classification/prediction, association rules2, Methods the survey questionnaire3, Customer behavior analysis 4, a variety of data mining algorithms: the TwoStep algorithm, K-Means algorithm, theQUEST algorithm, Apriori algorithm.5, data statistical analysis software: SASS, Clementine data mining platform.After data analysis, five conclusions are:1and a half years, online shopping more and each time shopping for between100-500much affected by online credibility with customers, but the proportion of the user is not high.Half year shopping a lot, every time shopping amount between100-500users affected byonline credibility to a certain extent, and the two types of user accounts for a large proportionin the group.2and a half years, online shopping more and online sales between100-500customersmore easily influenced by online search rankings, lower support showed lower proportion ofthe customers. It shows that most customers don’t influence by online search rankings.3, online shopping history for5years or more than5years, half a year online shoppingfrequency is higher; Less online shopping history within a year; Online shopping history in1-5years of two-thirds of the people shopping once in a while, a third of people often onlineshopping.4, the amount of online shopping and online shopping history is direct ratio relation,online shopping history long crowd consumption amount is rising trend, and taller; Onlineshopping history shorter people consumption amount is conservative; Online shopping historyfor1-5years of the crowd is more, consumption, tend to be moderate.5, online potential customers the crowd can be roughly divided into the followingcategories:(1) Internet access time in more than four years.(2) Internet users aged between18and35.(3) have college or bachelor degree of Internet users.In this article has collected the authoritative Internet survey and management institutions,on the analysis of the data used to validate the analysis conclusion, the results prove that thebasic accurately.更多还原