Researches On Contextualized Recommendation Model Oriented Customer Continuous Purchase Problem In E-Commerce Platform

With the development of mobile commerce, context-awareness and Internet of things, the boundary of e-commerce is greatly expanded, and it has entered a business information era called "big data". However, contradiction between disorderly, huge amounts of business information and the needs of customers is growing. In this situation, on the one the cost of e-business platform to acquire new customers is increasing sharply, on the other hand, the e-business enterprise want to retain existing customers and improve customer continuous purchase intention in order to maintain and improve the rate of return has become very urgent. E-business platform centered on B2C application has accumulated vast amounts of data, but customer facing the problem of "information resources are rich, but difficult to get useful information ". According to the customer’s preferences, historical network behavior and interest of other customer groups, etc., how to proactively provide goods to customer for their preferences, and to provide customers with personalized service, so as to stimulate customers continuous online browsing and buying behavior is the big challenges that face e-commerce platform.In recent years, personalized recommendation technology has received the widespread attention as an important way for customers to get preference goods in the vast business information. However, customer interest is complex in e-business platform, and purchasing behavior has become uncertainty and jump after affected by the contexts. Existing personalized recommendation service failed to deal with the above problems well and it cause loss of customers constantly. E-business platform has urgent needs to provide contextualized recommendation service for customer continuous purchase. The platform can accurately provide information services that both satisfy customer preference and meet the inside and outside contexts of customer.For this purpose, the research object of this paper is the problem of customer continuous purchase in e-commerce platform. The means is personalized recommendation methods affected by multi-dimensional context. The paper analyzes the diversity, personalized and dynamics of electricity customer context. And on this basis it takes decision-making behavior theory of network consumer, distributed cognition theory and maslow’s hierarchy of needs as the theoretical basis of personalized recommendation methods. In combination with methods such as clustering, decision tree, association rule, markov, collaborative filtering, ontology modeling and so on to research the solution for customers continuous purchase problem in e-business platform. This recommendation model proposed is applied to the different stage of customer’s continuous purchase in e-business platform. The main research work is as follows:1. Researches on problems of customer continuous purchase in e-commerce platform oriented contextualized recommendation modelCustomers continuous purchase problems in e-business platform can be categorized to two types:The first one is the old customer whose interests are stable in a certain period of time. For this type of customers, personalized recommendation model proposed obtains historical customer interest under multi-dimensional contexts, and recommends goods by using of the contextual recommended method; the second one is the old customer whose interests are drift due to the change of multi-dimensional context. It can be divided into incremental drift and radical drift. For this type of customers, personalized recommendation model proposed builds a dynamic interest model and continuous monitoring to adjust the changes in the customer interest, and finish push of the goods by using adaptive contextual recommended method. It analyzes e-business contexts and characteristic of customer interest in multi-level, multi-dimensional way. It creates the model that combines the context with customer’s interest as a knowledge support for contextualized recommended system.2. Research of contextualized recommended method based on the user’s sensitive contextsFor the continuous purchase problem of e-business platform customers whose interests are not drift away. However, traditional recommendation model does not consider the distributed and differentiated impact of different contexts on user needs, and it also lacks adaptive capacity of contextual recommendation service. Thus, a contextual information recommendation method based on user’s sensitive contexts is proposed. Firstly, the method analyzes the differential impact of various sensitive contexts and specific examples on user interest and designs a user interest extraction algorithm based on distributed cognition theory. Then, the sensitive contexts extracted from user are introduced into the process of collaborative filtering recommendation. The model calculates similarity among user interests. Finally, a novel collaborative filtering algorithm integrating with context and user similarity is designed.3. Research of contextualized recommended method that consider the customer interested in an incremental drift characteristicsFor the continuous purchase problem of e-business platform customers whose interests is in incremental drift characteristics. Firstly, it put forward the association rules algorithm based on modified FP-Tree, which effectively improves the mining efficiency of customer interest rule pattern in e-business platform. Secondly, it defines the context intensity, context correlation and they are processed in quantity. Based on this, it put forward users’interests mining algorithm integrating context intensity to model and express the customer interest under the influence of context contribution. And by using changes of association rule confidence and support, it does drift detection of contextualized customer preference pattern. Finally, in the collaborative filtering recommendation algorithm based on the item, it adopts the relationship between items in association rules for search of candidate items. Then it uses the context contribution which affects customer interest instead of score for dealing with the data of sparse and improves the accuracy of the similarity between items.4. Research of contextual ized recommended method that consider the customer interested in a radical drift characteristicsFor the continuous purchase problem of e-business platform customers whose interests is in radical drift characteristics. To address the interest drift problem brought from the difficulty that personalized information service is hard to effectively adapt to context and changes of user awareness, it propose a new method of contextual information recommendation. Firstly, user cognitive factors will lead to changes in interest; therefore, reasons of changes in user interest from the perspective of motivation are analyzed, and based on the Maslow hierarchy of needs, a mechanism is designed to analyze the information category and information behavior corresponding hierarchy of needs, and on this basis, a user interest determination algorithm is proposed based on ontology and hidden Markov. Secondly, it introduce the concept of user activity and present a user activity computational method integrated with context to solve the recommendation service of cold start and data sparsity problem. Finally, a dynamic collaborative filtering recommendation algorithm integrated with user activity is proposed, making the candidate recommendation content diversify selectively, by monitoring user feedback and learning interest variation, it can determine its interest trends and make the adaptation forwardly.5. Application on methods of customer continuous purchase in E-commerce platform oriented contextualized recommendation modelIt applies the method and model proposed in this paper to the continuous purchase problem in e-business platform. The paper designs system framework of recommendation system, and makes detailed analysis on its application. By using actual data in B2C platform, it verifies and analyzes the recommend effects of our contextual method in customer continuous purchase problem. It improve the quality of e-commerce personalized recommendation and put forward measures and suggestions for the intention of customer continuous purchase, which provides the reference for the research of the personalized recommendation in e-commerce enterprise application and customer retention.