Personalized Recommendation Of E-Commerce Based On Two-Stage

With the rapid development of Internet technology and swift growth of information resources,people gradually come into the age of "information overload".The"information overload" problem poses unprecedented challenges to information producers and consumers.As an important key to solve this problem,personalized recommender system has been developed rapidly and used widely in the mainstream e-commerce websites.In the e-commerce website,personalized recommendation plays a role of the sales staff,providing the services for users according to their needs,meanwhile,it can create more economic benefits by improving the user experience.In this paper,we proposed a two-stage e-commerce recommendation which contains score prediction and diversity recommendation,after analyzing the three problems of personalized recommendation of E-commerce.The lower accuracy of recommendation is solved by score prediction phase,which finally products list of initial recommended items.The problem of low excavation ability of long tail products is solved in diversity recommendation stage,so does compatibility between accuracy and diversity.In this stage,the initial recommendation list is filtered to produce the final Top-N recommendation list.We proposed the new scoring prediction algorithm based of NI-Slope One,which uses the improved method to calculate the similarity degree and find the nearest neighbor,and forecasts the unrated item in terms of the neighbor item's score after preprocessing.In the stage of diversity recommendation,we proposed a Top-N recommendation algorithm based on item grading re-calculation after analyzing the shortcomings of existing methods.Taking project popularity and user rating style into consideration,the algorithm recalculates the score of projects,and pick the proper individualized project in the initial list of recommended items for users according to the adjusted class weight and the final score.We designed experiment to verify the validity of proposed two-stage e-commerce recommendation.The experimental results show that the Top-N recommendation list generated by the proposed method has a great improvement in individual diversity and overall diversity in the condition of small loss of accuracy.It not only improves the user experience,but also creates more business value to the business by excavating long tail products for consumer.