Research And Application Of Collaborative Filtering Technology In Recommendation System

According to the data report of the Internet Center,even if we have a huge amount of information,the efficiency of finding useful information for ourselves is becoming lower and lower;this means that the contradiction between the massive supply of information and the needs of users is increasing.The recommendation system is a tool that helps users quickly find useful information,and is an effective solution to alleviate the contradiction between mass information supply and user needs.Among them,collaborative filtering technology is the most widely used technology in recommendation systems.The purpose of this paper is to solve the problems of sparse data,cold start,scalability and so on.Aiming at these problems,this paper puts forward its own solution,which mainly includes the following three aspects:(1)Similarity calculation is the core link of collaborative filtering technology.The accuracy of the calculation has a great influence on the recommendation result.In the case of very sparse data,in view of the large deviations in the commonly used similarity calculation results,which leads to unsatisfactory recommendation results,a new similarity calculation method is proposed in this paper.Based on the common scoring items,according to the closer the scoring time,the higher the similarity of the user,the proposed algorithm integrates the time factor;at the same time,in order to avoid the improper contribution of active users and popular items to the similarity calculation,this article also conducted related thinking.The experimental results show that the proposed method improves the accuracy of recommendation to a certain extent.(2)In order to reduce the neighborhood search space and improve the scalability of the recommendation system to a certain extent,this paper combines the K-means clustering algorithm with a collaborative filtering algorithm based on similarity calculation to make recommendations.Firstly,users are divided into corresponding classes or clusters by K-means clustering algorithm;then,other users with high similarity to the target user are found in each class or cluster by combining the similarity calculation method in Chapter 3;finally,the nearest neighbor with high similarity is selected to form the target user's nearest neighbor,which is predicted and recommended according to the nearest neighbor's score.At the same time,in order to highlight the significance of users' recent ratings for recommendation results,this paperintroduces a time decay function in rating prediction.Experimental results show that the algorithm is feasible to some extent.(3)This paper studies the movie prototype recommendation system based on collaborative filtering,gives the development environment and implementation interface,and analyzes and discusses the architecture and core module design of the system.