Effect Of The Diversity Of Personalized Recommendation On Recommendation Efficiency In Electronic Commerce Websites

With the rapid development of Internet, the network has become not only a tool to connect different regions and groups, but also an indispensable way to obtain information and shop online. Especially in twenty-first century, e-commerce has attracted significant attention around the world, and the mode that buyers and sellers complete the transaction through network instead of meeting goes through a swift rise driven by information and technology. However, along with the increase of network information,it has been difficult for consumers to browse all the products in a short period due to massive information on e-commerce websites, which will eventually result in information overload, under this circumstances, personalized recommendation system arises at the historic moment. Personalized recommendation has. an important impact on alleviating information overload, improving the efficiency of online shopping,and promoting product sales. Nevertheless, even though e-commerce websites at home and abroad have adopted a variety of personalized recommendations to improve the user experience,consumers still face the dilemma of information overload due to inappropriate recommendations. How to improve the effect of personalized recommendation has been widely concerned by researchers and e-commerce website managers.Most existing studies focus on how to improve the accuracy and efficiency of the recommendation algorithms, or are more concerned on ways to reduce perceived risks and thus increase consumer satisfaction. However, few studies have focused on or explore how the diversity of personalized recommendations will affect the effectiveness of the recommendation efficiency based on the process of consumer decision making in the personalized recommendation process. In view of the current research gaps and deficiencies, our study begins from the consumer decision-making process, using consumers' two-stage decision-making theory,preference inconsistency theory,long tail theory and anchoring effect as a basis. Based on the theories and methods of management. marketing, computer science and other disciplines, this paper makes an empirical analysis and modeling of the effect of personalized recommendation diversity on recommendation efficiency in e-commerce websites. The results of the empirical analysis in this paper are consistent with the theoretical prediction results, and have certain scientific research value and practical value.The main conclusions and achievements of this paper are as follows:First, this paper established the theoretical model of the effects of recommendation timing and product portfolios in the efficiency of personalized recommendation, studied the effect of recommending different products portfolio to consumers in different times on recommendation efficiency.Based on the two-stage decision-making theory and inconsistent preference theory,this paper divided the recommendation timing into two stages - first stage, when users browse products, second stage, before users click their final choices. In addition, this paper divided the recommended products into similair products and related products,combined the recommendation timing with recommended product portfolio to establish a theoretical model of the effects of recommendation timing and product portfolios in the efficiency of personalized recommendation, and studying the effect of recommending different products portfolio to consumers in different times on recommendation efficiency. Research shows that in the first stage of decision-making,consumers tend to prefer more variety of choices,thus, compared to the recommendation received in the second stage, consumers tend the accept product recommendations they received in the first stage. The results of the study also show that during the process of making consideration set by consumers, their focus is not on the similar products, but more on diversely recommended products. Moreover, during the second stage of the decision-making process, as we had predicted, consumers focus shift to products complementary to the target product. Consumers start to give attention to complementary product recommendations and consider buying products other than the original products they were planning on purchasing.Second, this paper established the theoretical model of the effect of sales and scores of recommended product in the efficiency of personalized recommendation,and studied the independent and interactive influence of the two factors on the recommendation efficiency.In this paper, the recommended products were divided into mainstream products and niche products. We selected one form of product reviews---product score combined with product sale,established the theoretical model of the effect of sales and scores of recommended product in the efficiency of personalized recommendation. The conclusions are: first, compared with just recommending the mainstream products or the niche products, consumers will adopt more recommended products when the system recommended both mainstream products and niche products at the same time. Second,compared with just recommending the high score products or the low score products,the recommendation efficiency doesn't get improved, and consumers are not tend to purchase more recommended products. Third, compared with just recommending the high score mainstream products or the low score niche products, consumers are more likely to accept the recommended product when the two kinds of products are recommended at the same time. Similarly, compared with just recommending the high score niche products or low score mainstream products,the result is still the same when they are recommended at once.Third, this paper divided the recommend products into several different types when establishing the two theoretical models. This paper further explored differences in the acceptance of personalized recommendations among different types of products, to promote the research of personalized recommendation diversity, from the perspective of management and consumer behavior.From the perspective of two different disciplines between marketing and information economics, this paper divided the recommended products into several different types with recognized product classification, such as hedonic products and practical products,search products and experience products. When establishing the theoretical model of the effects of recommendation timing and product portfolios in the efficiency of personalized recommendation, this paper used hedonic products and practical products as moderator variables. Similarly, when establishing the theoretical model of the effect of sales and scores of recommended product in the efficiency of personalized recommendation, this paper used search products and experience products as moderator variables. On this basis, this paper further explored differences in the acceptance of personalized recommendations between different types of products, and discovers that recommendations of hedonic products are more effective than that of practical products, experience products are more effective than that of search products.The theoretical contribution of the study includes the follow three aspects.First, this paper begins from the consumer decision-making process,reveals the consumers' acceptance mechanism of personalized recommendations. Second, the paper explores the effect of the personalized recommendation diversity on recommendation efficiency, verifies that consumers demand diversity in recommended content. Third, this paper explores the effect of different product categories on the recommendation efficiency, and reveals differences in the acceptance of personalized recommendations between different types of products.The practical implications of the study includes the follow tow aspects.First, when designing a personalized recommendation system,one should take full account of characteristics of consumers' behaviors. When consumers are forming a consideration set, a personalized recommendation system should provide the consumers with more choices. In addition, e-commerce sites should take advantage of consumers strong preference towards hedonic products and experience products, by promoting sales of hedonic products and experience products through personalized recommendations. For practical products and search products, in addition to continuing to make improvement on recommendation algorithm, there should be further exploration on ways to combine it with other marketing tools to promote product sales.