Research Of Product Personalized Recommendation Based On Virtual Communities

With the continuous development of network marketing, especially emergence of online group-buying which is a new e-commerce model, demand difference between consumers becomes more and more obvious. Therefore, mining demand preferences of individual consumers and providing them personalized products becomes an important way in marketing businesses. Because of the convenience of information dissemination consumer purchasing decision is not only just influenced by personal preferences, but also by views of others in a virtual community. How to take these two factors on consumer purchase intention into consideration and provide consumers more accurate personalized products becomes a hot issue.Based on product attributes and influence of opinions of others, a conceptual recommendation model is put forward in this study. There are32product attributes from7categories and opinion leadership value which in it. Thus, the final model is established by means of BP neural network.In this study,100users in a university forum are surveyed about their purchasing intention on35products. Prediction for their willingness to purchase these products is obtained by the research model and compared with actual data. Based on results of the empirical analysis, a few suggestions are proposed and following conclusions are obtained.1. The conceptual model in this study can be used to describe the factors affecting consumer willingness to buy online product, especially for consumers who are active in a virtual community. In virtual communities, consumers purchasing decisions are not only affected by their preference for product attributes, but also by the views of others.2. Opinion leader value can reflect the influence of one person’s view on other consumers’purchase decisions.3. The algorithm proposed in this study to identify the opinion leaders in virtual communities has a higher degree of accuracy and credibility. Compared to the traditional method based on the number of references, this algorithm takes into account the three factors of the number of references, user’s influence and degree of dependence on other users, thus makes the evaluation more objective. Besides, it reduces the influence of online water army.