Research On Personalized Recommendation In E-commerce

The rapid development of Internet technology has taken the human society into the Information Age. E-commerce as a new business model also quickly spreads. Emergence of electronic commerce has changed the traditional consumption model and brought our lives great convenience. As E-commerce grows and provides more and more goods, so that it makes consumers hard to find products to interest and lots of information without visiting becomes "dark information". In this case, a variety of recommendation systems came into being, the function of what is to provide users with the appropriate recommendation.We choose e-commerce site as the research object, we proposed a SPS (a set of similar products) algorithm based on the category system of e-commerce, the method can overcome the difficulty of e-commerce such as product classification inaccuracy and so on. Finally, we describe the detail of technique and the algorithm. Then, this article proposed a personalized recommendation SPS-based algorithm based on the research of recommendation system of e-commerce, the method can mining the "dark information", and has high accuracy, it's appropriate to apply in the e-commerce website; Finally, this paper given technical solutions, and the whole system technology.The main research contents and characters of this paper are as follows:1) Design and implement an automated category building system. First, we analysis the feature of existing e-commerce site and products, we then used user behavior information for mining the important attribute of products. Finally, we make use of the properties of these important dimensions and other information of products, and text mining algorithms to clustering the products.2) Design and implement a personalized recommendation based SPS system. Traditional collaborative filtering algorithms are widely used in e-commerce, with a high recommendation accuracy rate, but it can not solve the problem of sparseness of user behavior, in practice often recommended to bring the long tail problem. We improve the traditional collaborative filtering algorithms by design the SPS-based algorithm, the experimental results shows that the method we proposed have high coverage and have a good performance in recommend new products.