SIA OpenIR  > 数字工厂研究室
面向O2O的潜在客户识别的研究与应用
Alternative TitleResearch and Application of Potential Customers for O2O Marketing
翟鹏华
Department数字工厂研究室
Thesis Advisor张丁一
Keyword潜在客户 O2O 随机森林 加权投票
Pages63页
Degree Discipline控制工程
Degree Name硕士
2019-05-17
Degree Grantor中国科学院沈阳自动化研究所
Place of Conferral沈阳
Abstract随着当今社会的信息化程度越来越高,电子商务行业蓬勃发展,每天都会有大量的用户在电子商务网站中浏览商品或者进行购物等,因此在电子商务网站中每天都会有大量的用户行为信息被保存下来,这些数据不仅能够反映出用户当前的浏览行为,更重要的是能够反映出每个用户潜在的购买意愿。因此,如果能够从这些数据中准确高效的挖掘出电子商务企业的潜在客户,商家就可以针对这些客户进行个性化的服务,实现精准营销,能够最大程度地将潜在客户转化成实际客户,从而获得更多的利润,最终在竞争激烈的电子商务市场竞争中占据有利地位。O2O营销模式能够充分的利用线上线下的资源,经营方式多样化,用户能够在线上进行咨询或者支付等行为,并在线下进行消费,能够充分满足用户的需求,是一种非常重要且受欢迎的电子商务营销模式。因此,结合O2O营销模式以及电子商务企业的实际需要,本文的主要任务就是研究如何从庞大的O2O电子商务数据中挖掘出商家的潜在客户,以便商家向用户提供个性化的服务,实现精准营销,提高O2O电子商务企业的市场竞争力。本文的主要研究工作如下:首先,进行了大量的调研分析。探讨了潜在客户的特点、研究现状以及本课题的研究意义,分析了现有的潜在客户识别的国内外研究现状与成果,发现现有研究的不足,掌握国内外的研究进展。其次,构建了潜在客户识别有效特征集。因为O2O电子商务数据具有高维性和海量性的特点,并且在这些数据中还存在大量的冗余数据和噪声数据。因此,本文对数据进行了预处理,将数据转化成可处理的结构化数据,并利用这些数据进行潜在客户识别有效特征集的构建,以便建模对潜在客户进行挖掘。再次,潜在客户识别模型的构建。随机森林是一种非常有效的分类预测算法,具有很高的预测精度。并且异常值和噪声对于算法的影响很小,具有很好的泛化能力。在这部分通过分析电子商务数据的特点以及结合现有算法的优缺点,提出了一种改进的随机森林算法,建立了面向O2O的电子商务潜在客户挖掘模型。最后,实验验证与结果分析。利用常用的公共数据集以及某O2O电子商务企业的数据对构建好的潜在客户挖掘模型进行实验验证和分析。试验结果表明,改进后的算法的预测准确率有了一定程度的提升。之后根据数据的特点设计了一种模型的评价指标,并在潜在客户数据集上验证了该评价指标的稳健性,能够很好的评价模型的预测能力。实验结果表明,构建好的潜在客户识别模型能够很好的拟合电子商务数据,并较为准确的挖掘出网站中的潜在客户。
Other AbstractWith the improvement of the degree of informatization in today's society, the e-commerce industry is booming, and every day there are a large number of users browsing goods or shopping on e-commerce websites. Therefore, a large amount of users’ behavior information is saved every day in e-commerce websites. These data not only reflects the users’ current browsing behavior, but more importantly reflects the potential purchase intention of each user. Therefore, if we can extract potential customers of e-commerce enterprises accurately and efficiently, merchants can carry out personalizing services for these potential customers and achieve precise marketing. And merchants can convert potential customers into the actual customers of the business to the greatest extent, in order to gain more profits. Ultimately, merchants will occupy a favorable position in the highly competitive e-commerce market. O2O is a kind of very effective and popular e-commerce marketing model. O2O can make full use of online and offline resources, and diversify its business methods. In the O2O marketing model, customers can conduct online consultation or payment, and then they can consume offline, and merchants can fully meet the needs of users. So, O2O is an important e-commerce marketing model. Therefore, combining O2O marketing model with the actual needs of e-commerce enterprises, the main task of this paper is to explore how to dig out potential customers from huge O2O e-commerce industry data, so that merchants can provide personalized services to customers and achieve precision marketing, in order to improve the market competitiveness of O2O e-commerce companies. The main research works of this paper are as follows: Firstly, conducting a lot of basic theoretical analysis. Discussing the characteristics of the potential customers, the research status and the research significance of this subject. And we analyze the current research status and achievements of existing potential customers identification to discover the shortcomings of existing researches and grasp the research progress at home and abroad. Secondly, constructing a potential customers identification valid features set. Because O2O e-commerce data are characterized by high dimensionality and massiveness, and there are a large amount of redundant data and noise data in these data. Therefore, we have preprocessed these data in this paper, transforming these data into structured data. And we use these data to construct the valid feature set of potential customers identification, in order to construct model to mine potential customers from these data. Next, constructing the potential customers identification model. Random forest is a very effective classification algorithm with high prediction accuracy. And unusual values and noise values have little effects on random forest algorithm, and it has good generalization ability. In this part, by analyzing the characteristics of e-commerce data and combining the advantages and disadvantages of existing algorithms, we propose an improved random forests algorithm, and establish an e-commerce potential customer mining model for O2O. Finally, conducting experimental verification and results analysis. Experiments and analyzes the constructed potential customer mining model by using common public data and the data from an O2O e-commerce enterprise. The experimental results show that the prediction accuracy of the improved algorithm has been improved to some extent. Then, according to the characteristics of data, a model evaluation index is designed. And the robustness of the evaluation index is verified on the potential customers dataset, which can well evaluate the prediction ability of model. And the experimental results show that established potential customer identification model can well fit e-commerce data and mine the potential customers from e-commerce websites accurately.
Language中文
Contribution Rank1
Document Type学位论文
Identifierhttp://ir.sia.cn/handle/173321/25196
Collection数字工厂研究室
Recommended Citation
GB/T 7714
翟鹏华. 面向O2O的潜在客户识别的研究与应用[D]. 沈阳. 中国科学院沈阳自动化研究所,2019.
Files in This Item:
File Name/Size DocType Version Access License
面向O2O的潜在客户识别的研究与应用.p(1128KB)学位论文 开放获取CC BY-NC-SAApplication Full Text
Related Services
Recommend this item
Bookmark
Usage statistics
Export to Endnote
Google Scholar
Similar articles in Google Scholar
[翟鹏华]'s Articles
Baidu academic
Similar articles in Baidu academic
[翟鹏华]'s Articles
Bing Scholar
Similar articles in Bing Scholar
[翟鹏华]'s Articles
Terms of Use
No data!
Social Bookmark/Share
All comments (0)
No comment.
 

Items in the repository are protected by copyright, with all rights reserved, unless otherwise indicated.