Research On Personalized Recommendation Method Based On Dynamic Evolution Mechanism Of Consumption Preference

With the rapid development of network information technology, e-commerce is growing fast in recent years, and online shopping has become an important part of people’s network of life. The rapid development of online shopping and lots of products flooding to the e-commerce network are bound to the "information overload" problem. Recommended system is an effective means to solve this problem. Collaborative filtering recommendation system technology is the most widely used and the most effective technology, but it faces the sparsity of user data, cold start, scalability and other problems. As e-commerce system integrated social networking services, it provides the basis for the integration of social networks with social network analysis in the recommended technique. Meanwhile, the integration has improved some of the problems that exist in the recommended techniques effectively. Therefore, research on how to apply social network analysis in personalized recommendation system is not only can improve the common issues but also can provide users with more accurate and effective recommendation. Thereby it can increase the customer’s satisfaction and purchase rate, which is of great practical significance to users and enterprises.The paper get user’s consumption preferences situation based on analysis of users’ historical behavior data. Binding the study on information dissemination in social networks to analyze the impact of user consumption by friends, and then update user’s preferences for the latest. Last, it can obtain the target product to recommend to the user by the match of the latest consumer preferences and commodity attributes documentation.First, under this paper’s research background, describe commodity attributes relationship with qualitative and quantitative characterization of user’s historical commodity data by studying the literature on consumer preferences at home and abroad. Analyze the preference degree that user preference for feature item based on the vector space model to build user consumer preference model.Second, define friends’ relationships intensity with interactions between users in the social network and analyze user’s consumer preference’s propagation process based on the improvement of Linear Threshold Model. Then construct user consumption preferences’ propagation model by research the dynamic process of user consumption preference and based on the dynamic process to define the dynamic evolution mechanism of user preferences. Update user consumption preference if the user consumption preferences’ propagation model has an impact on it.Third, apply the dynamic evolution mechanism of user preferences into the personalized recommendation and design the matching rules of user consumption preference and commodity attributes. According to the user’s latest consumption preference, recommend the target product to the user based on the matching rules. Compared to the traditional content-based recommendation algorithm, the proposed recommendation method has more accuracy and efficient on the recommended results.What’s more, the innovation of this paper are the analysis of consumer preferences by mining user historical behavior commodity data and the analysis of propagation mechanisms of consumer preferences based on the interactions between users in social network. According to the research, it provides users with more accurate goods and improve customer satisfaction and business efficiency. Study on consumer preferences propagation mechanism can not only help us understand the impact effect, also can predict the trend of consumer preferences and recommended product to the user more accurately in advance that he might be interested.In addition, the basic data of this study is the user’s historical behavioral data and interactive information between users. But these basic data involve the business enterprises’ confidential that can’t get the user’s actual e-commerce system behavior data. Due to this limitation, using the simulation method to access to the research data by building an e-commerce simulation systems in this paper. In order to ensure the data reliability, allow users to browse, shopping, collection, forwarding, interactive action and other acts in the simulation system by adopt certain incentives. Meanwhile, using the R language to achieve the validation of the models and recommendation algorithm with R software platform.