电子商务网站个性化推荐的多样性对推荐效果的影响研究

发布时间：2018-05-24 02:50

本文选题：个性化推荐 + 推荐多样性　；参考：《北京邮电大学》2017年博士论文

【摘要】：随着Internet的迅猛发展,互联网在人们生活中所扮演的角色不再只是连通不同地域、不同人群的工具,网络已经成为人们获取信息、购物消费的不可或缺的方式。尤其是进入21世纪后,电子商务在全世界范围受到了高度的重视,这种买卖双方不谋面、通过互联网进行交易的模式在信息技术的带动下迅速崛起。然而,随着网络信息资源的规模不断扩大,消费者面对电子商务网站提供的海量信息时,很难做到在短时间内浏览所有的产品,最终将导致信息过载,因此个性化推荐系统应运而生。个性化推荐对缓解信息过载、提升网络购物效率、促进产品销售具有重要影响。然而,国内外电子商务网站虽然采用了多种个性化的推荐形式来改进用户体验,但是不恰当的推荐使消费者依然面临信息过载的困境。如何提升个性化推荐的效果,受到了研究人员和电子商务网站管理者的广泛关注。既有研究多数关注如何提高个性化推荐算法的精确度,或者关注如何降低消费者的感知风险进而提高满意度。然而,鲜有研究基于消费者的决策过程来关注和探究个性化推荐过程中推荐的多样性对推荐效果会产生如何的影响。针对先前研究中存在的不足,本文从消费者的决策过程出发,基于消费者两阶段决策理论、偏好不一致悖论、长尾理论以及锚定效应等理论,综合运用管理学、营销学、计算机科学等学科的思想和方法,对电子商务网站中个性化推荐的多样性对推荐效果的影响进行了实证分析和建模研究。本文的实证分析结果与理论预测结果基本一致,具有一定的科研价值与实践价值。本文的主要研究结论和研究成果如下:第一、建立了推荐时机与推荐产品组合对推荐效果影响的理论模型,研究了不同时机下向消费者推荐不同的产品组合对个性化推荐效果的影响。本文基于消费者两阶段决策理论与偏好不一致悖论相应的将推荐时机划分为两个阶段,即形成考虑集合和做出最终选择前后两个阶段。此外,本文将所推荐产品的划分为同类产品与相关产品来进行研究,并结合推荐时机与产品组合两个因素构建了推荐时机与推荐产品组合对推荐效果影响的理论模型,研究了不同时机下向消费者推荐不同的产品组合对个性化推荐效果的影响。得到的结论有:(1)与在决策第一阶段收到产品推荐相比,消费者更倾向于采纳决策第二阶段收到的产品推荐;(2)在考虑集合形成阶段中,消费者的关注对象并没有明显集中于同类产品,而对推荐产品的多样性的诉求更为突出;(3)在最终做出消费决策时,消费者开始关注相关产品的推荐,考虑目标产品以外的购买。第二、建立了推荐产品销量与推荐产品评分对推荐效果影响的理论模型,研究了推荐产品的销量与评分两个因素单独作用下和交互作用下对推荐效果的影响。本文将推荐的产品划分为主流产品与利基产品来进行研究,并选取了产品评论的其中一种形式——产品的评分结合产品的销量构建了推荐产品的销量与推荐产品评分对推荐效果影响的理论模型,分别探究了这两个因素在不同的推荐情景中单独作用或交互作用下对推荐效果产生的影响。得到的结论有:(1)与仅推荐主流产品或仅推荐利基产品相比,当系统同时推荐主流与利基产品时,消费者会采纳更多的推荐产品;(2)与仅推荐高评分产品或仅推荐低评分产品相比,当系统同时推荐这两种产品时,并没有改善推荐效果,消费者并不会多的去购买系统所推荐的产品;(3)对比系统仅推荐高评分主流产品或仅推荐低评分利基产品的情况,当系统同时推荐二者时,消费者更倾向于采纳系统所推荐的产品;而对比系统仅推荐高评分利基产品或仅推荐低评分主流产品的情况,当系统同时推荐二者时,消费者更倾向于采纳系统所推荐的产品。第三、在构建两个理论模型的同时,相应的将推荐产品划分不同的类型,进一步探索了消费者对不同类型的产品推荐采纳的差异,推动了管理与消费者行为视角下的个性化推荐多样性的研究。本文分别从营销学与信息经济学两个不同的学科视角,以公认的产品分类方式对应的将推荐的产品划分为享乐品和实用品、搜索品和体验品等多个不同的类型。在构建推荐时机与产品组合对推荐效果影响的理论模型时将享乐品和实用品作为调节变量,同时在构建推荐产品销量与评分对推荐效果影响的理论模型时将搜索品和体验品作为调节变量,进一步探索了消费者对不同类型的产品个性化推荐的采纳差异。研究结果显示,消费者对不同类型的产品存在推荐采纳的差异,享乐品的推荐效果好于实用品的推荐效果、体验品的推荐效果好于搜索品的推荐效果。本文的理论贡献主要体现在以下三个方面:第一,本文从消费者的决策过程出发,揭示了消费者对个性化推荐的采纳机制;第二,本文探索了个性化推荐的多样性对推荐效果的影响,验证了消费者对推荐内容多样性的需求;第三,本文进一步探究了不同的产品分类对推荐效果产生的影响,揭示了消费者对不同类型产品个性化推荐的采纳差异。本文对实践的启示意义包括以下两个方面:第一,本文强调应充分考虑消费者在不同决策阶段的行为特点来改善个性化推荐系统的设计,在考虑推荐精确性的同时充分重视推荐内容的多样性;第二,本文强调应充分利用消费者对享乐品和体验品推荐内容的强偏好性,通过个性化推荐来促进这两种产品的销售,而对于实用品和搜索品,除了继续完善推荐算法外,还需进一步探索与其他营销手段相结合来促进产品的销售。
[Abstract]:With the rapid development of Internet, the role of Internet in people's life is no longer only a way to connect different regions and different groups of tools. The network has become an indispensable way for people to obtain information and purchase consumption. Especially after entering twenty-first Century, e-commerce has been highly valued all over the world. However, with the increasing scale of the network information resources, it is difficult for consumers to browse all the products in a short time as the scale of the network information resources is expanding, and it will eventually lead to information overload, so personalized recommendation. The personalized recommendation has an important impact on alleviating the information overload, improving the network shopping efficiency and promoting the product sales. However, the domestic and foreign e-commerce websites have adopted a variety of personalized recommendation forms to improve the user experience, but the inappropriate recommendation makes the consumers still face the difficult situation of information overload. The effect of the promotion of personalized recommendation has attracted wide attention from researchers and E - commerce website managers. Most of the research focuses on how to improve the accuracy of personalized recommendation algorithms or how to reduce the perceived risk of consumers to improve their satisfaction. According to the shortcomings of the previous research, this paper, based on the consumer's decision-making process, based on the consumer's two stage decision theory, the preference inconsistency paradox, the long tail theory and the anchoring effect, makes a comprehensive use of management, marketing, and calculation. The influence of the diversity of personalized recommendation on the recommendation effect in e-commerce websites is analyzed and modeled. The results of this paper are basically consistent with the theoretical prediction results, and have certain scientific research value and practical value. The main research conclusions and research results of this paper are as follows. First, the theoretical model of the effect of recommendation time and recommended product combination on the effect of recommendation is set up, and the effect of recommending different product combinations to consumers at different times is studied. This paper divides the recommendation time into two stages based on the corresponding two stage decision theory of consumers and the inconsistent paradox of preference. In addition, this paper divides the recommended products into two stages. In addition, this paper divides the recommended products into similar products and related products, and constructs a theoretical model of the effect of recommendation timing and recommended product combination on recommended effects combined with two factors of recommendation timing and product combination, and studies the different opportunities at different times. The effect of different product combinations on Personalized Recommendation effect is recommended to consumers. The conclusions are as follows: (1) consumers are more inclined to adopt product recommendation in the second stage of decision making than to receive product recommendation in the first stage of decision making; (2) in consideration of the formation stage, the consumer's concern is not obviously concentrated in the same category. The demand for the diversity of recommended products is more prominent; (3) in the end of the decision making, consumers begin to pay attention to the recommendation of the related products and consider the purchase outside the target product. Second, a theoretical model of the effect of the recommended product sales and the recommended product score on the recommendation effect is established, and the sales and evaluation of the recommended products are studied. This paper divides the recommended products into the mainstream and niche products, and selects one of the forms of product reviews - the product score and the sales volume of the product to construct the recommended product sales and recommended product scores on the recommendation effect. Two The theoretical model of fruit impact examines the effects of these two factors separately or interacting on the different recommended scenarios. The conclusions are as follows: (1) consumers will adopt more recommendations when the system recommends the main stream and niche products at the same time compared with only the mainstream products or only the niche products recommended. Products; (2) when compared with only highly rated products or only recommended low grade products, when the system recommends these two products at the same time, it does not improve the recommendation effect. Consumers will not be able to buy more products recommended by the system; (3) the contrast system recommends only high grade mainstream products or only low score niche products, when the system is the same When the two are recommended, consumers are more inclined to adopt the products recommended by the system; and the contrast system recommends only high grade niche products or only the low score mainstream products. When the system recommends the two, the consumer is more inclined to adopt the products recommended by the system. Third, while constructing the two theoretical models, the corresponding will be Recommending different types of products, further exploring the differences between consumers and different types of products recommended by consumers, and promoting the diversity of personalized recommendation from the perspective of management and consumer behavior. This paper, from two different disciplinary perspectives of marketing and information economics, will correspond to recognized product classification methods. The recommended products are divided into hedonic and practical products, search products and experiential products, such as many different types. The results show that consumers have recommended differences in different types of products, and that the recommendation effect of pleasure products is better than that of practical products. The recommendation effect of experiential products is better than that of search products. The theoretical contribution of this article is mainly embodied in the following three aspects: first, this paper, starting from the decision-making process of the consumer, reveals the adoption mechanism of the consumer's personalized recommendation; second, this article explores the influence of the diversity of personalized recommendation on the recommendation effect, and validating the consumer's demand for the diversity of the recommended content; third, This paper further explores the impact of different product classifications on the effect of recommendation effect, and reveals the adoption differences of consumers' personalized recommendation for different types of products. The implications for practice include the following two aspects: first, this article emphasizes that the behavior characteristics of consumers at different decision-making stages should be fully considered to improve individualization. Recommending the design of the system to give full attention to the diversity of the recommended content while considering the accuracy of the recommendation. Second, this article emphasizes that it should make full use of the strong preference of the consumer for the content of the pleasure and experience, and promote the sales of the two products by personalized recommendation, and to the practical and search products, in addition to continuing to perfect the recommendation. The algorithm also needs further exploration and other marketing means to promote product sales.
【学位授予单位】：北京邮电大学
【学位级别】：博士
【学位授予年份】：2017
【分类号】：F724.6;F274

【相似文献】