E-commerce Personalized Recommendation Based On Hybrid Mechanism Learning

With the rapid development of network technology,E-commerce development is becoming increasingly popular and prevalent,with an increase in the amount of information,goods,information overload is grim increasingly,so it need to create a personalized recommendation system of E-commerce services,users will truly meet demand for goods to the customers and increase the sales force to Web site.Recommended personalized recommendation algorithm is the core of the system, at present,the application of collaborative filtering algorithm is the most popular recommended method,but there are still many deficiencies in practical applications: such as the score matrix sparse problems,accuracy problems,In this paper,It has proposed a hybrid mechanism based on personalized recommendation technology,not only solution to the issue of the score matrix sparse effectively,but also enhance the accuracy of the recommendation.The main research results are as follows:(1) By use of the user's registration information,cluster users with improved the K-means clustering algorithm,assigned to a cluster in the same concentration of similar personality or preferences,so that,you can find in the cluster in the user's nearest neighbor to avoid the entire user base in the search and improve the real-time response speed,and indirectly,the beginning of the solution to the problem of.The first evaluation.(2) The attributes of the item is divided into m dimensions,each dimension has its own characteristic value of the property,then create the eigenvalue matrix,by calculating the properties of similar characteristics between the two Item,and get a rough score to enrich the score matrix,for the subsequent calculation of the similarity between users,work ready.(3) The recommendation of the traditional algorithm does not take into account the categorization of item which the user preferences,this paper presents the project to calculate the user's preference value category,the combination of the recommendation to enhance the accuracy of the recommendation.(4) Finally design a personalized e-commerce recommendation system framework based on the algorithm proposed in this paper,It has a detailed description of the recommended processes,this framework can be applied to audio-visual,cultural and other recommended products,but also large-scale e-commerce systems can be used as a sub-module to run.