Research And Implementation Of Shopping Website Recommendation System Based On Collaborative Filtering

The development of e-commerce has brought development opportunities for the personalized recommendation system,in which various historical information and behavior are analyzed and some useful information is filtered for users among the massive information.With the gradual improvement of the requirements of the users,more effort is devoted to solving the issues in existing researches.Because of the growing of the number of the users,the products and the losing of scoring data,the data sparsity problem has been more and more prominent and the accuracy of the recommendation method has degraded significantly.To address this,an improved collaborative filtering recommend algorithm based on binary K-means algorithm is proposed.To solve the cold start problem in the algorithm,a proportional mean predictive scoring method is proposed to improve the performance of the algorithm.Finally,this thesis adopts the two methods above to design and implement a recommendation system for shopping website.The main contents of this thesis are as follows:(1)A collaborative filtering recommendation algorithm based on the improved binary K-means is proposed to solve the sparse data problem in the collaborative filtering recommendation algorithm.This thesis uses the improved binary K-means to cluster witch can reduce the effect of the choose of the initial centroid on the clustering result by adding the centroid dynamically.In order to reduce the differences between the users' standards,we put the average value and the similarity of users in the same cluster together to calculate predictive scores.The results show that the algorithm in this thesis is better.(2)In order to solve the cold start problem in the collaborative filtering recommendation algorithm,a proportional mean predictive scoring method is proposed to further improve the method.The algorithm uses the proportional mean predictive scoring to correlate the users' scores with the users' attributes effectively.By using the registration information of new users and new products to improve the cold start problem.By comparing with the mean method and the mode method on the different scales of data,it shows that the method in this thesis has a better result.(3)By analyzing the requirements,functional modules and processes of the shopping website,this thesis combines the improved binary K-means method and the proportional mean predictive scoring method to design a recommendation system for shopping website based on collaborative filtering which can improve the data sparse problem and the cold start problem.(4)By analyzing and designing the function of the system,this thesis implements the recommendation system for shopping site based on collaborative filtering.The system uses the SSH framework technology to separate the data display,the data storage and the control management.At the same time,we select the W test model and use the LoadRunner tool to test and verify the system.The test result shows that compared with the traditional recommendation system,the quality of the system is improved effectively with better performance and better stability.