Based On Collaborative Filtering Of E-commerce Personalized Recommendation System Realization

Nowadays, E-commerce, which has an important form as shopping based on Internet, has been developed rapidly. The personalized recommender system can improve the convenience and efficiency for shopping and becomes the hot research domain. The kernel of the recommender system is the adopted recommender algorithm. The cooperative filtering method is one of the simple and well recommender algorithms. This paper designed and reliazed a personalized recommender system based on the cooperative filtering method, taking the filem recommendation as an example. The researched system can recommend the appropriate filems corresponding to the preference of users, void the eysless purchasing process and reach the aim of preserving users and improving saling efficiency.The main contributions of this paper are listed as follows:(1) A cooperatative filtering algorithm is designed and reliazed. The idea of the cooperative filtering algorithm is introduced and the addressed indexes are listed. An algorithm flow has been designed, including data pretreatment, the clutering of users and items, the calculatation of the similarity indexes, and the selection of Top-N nearest neighbors; according to the matrix character of the user-item rating value, the Matlab softerware is adopted to realize the designed algorithm to reduce the programming load and improve the calculation efficiency; it analyzed the performance the designed algorithm, based on the MovieLens dataset, and compared the effect on the different clustering number, similarity indexes and N value on the recommendation performce. The experiment results illustrated the validity and robust of the designed algorithm.(2) The recommender system is desiged.The modules of the software and the modules'relations are designed by analyzing the functions and requirement of the recommeder system, and the developing envirment and the software form is fixed on; the widly-usered Access software is adopted to design the datebases and storage the required data, inluding the user information, the iterm information and the rated values; it desiged the appearance and the windows of the software, and the detailed function of each window and controls were presented.(3) The recommender system, based on the method of cooperate filtering is realizedThe main interface of the system, the scheme for realizing the system and the accessing mode of the databses is given. The software was developmented based on the envirment named as Visua Studio.net. The recommender system modules are realized via the fixed promgramming form of C# abd Matlab, which reduced the programming load; the variables and functions of eaech windows are designed and realized, and the processing workflow of important functions are presented.