The Comparative Study Of The Algorithm Based On Collaborative Filtering Recommendation System

With the continuous development of information technology, the Collaborative Filtering Recommendation technology as of the most important one of the recommended techniques currently even in the future, is an important research aspect of E-Commerce System. There are some disadvantages forced to be improved in the Traditional Filtering Recommendation system. For example:the data sparsity, cold start, the impact of the recommendation due to the time difference.The subject has a more in-depth exploration of the definition of the concept, the content and the elements of the recommendation system, and a more detailed description of the recommended techniques, comparative analysis of the advantages and disadvantages of these recommended techniques. On the basis of this, the thesis lays emphasis on the research of collaborative filtering recommendation technology, analyses and researches the problems of the correlation algorithm, and explores the improvement ideas of the current collaborative filtering algorithm. The contents mainly include:Firstly the thesis analyses the existing recommended technical methods, and compares and summarizes the shortcomings and defects in these technologies.Secondly the thesis explores the main problems of the collaborative filtering recommendation technology, and explored in detail the implications of the data sparsity and time, compares the similarities and differences of the User-Based and Item-Based collaborative filtering algorithms, and introduces the common similarity calculation methods:cosine similarity, modified cosine similarity and related similarity.Then the thesis introduces the related contents of the Cloud Model, and joins the Cloud Model into the collaborative filtering recommendation algorithm; use the similarity of the Cloud Model instead of the previous vector similarity to improve the accuracy of the algorithm. Toughing further summary and research on two aspects of the traditional algorithm, this research has combined user’s rate for item, user’s preference for item’s characteristic attributes, time effect into an effective unity, made the effective improvement of the traditional collaborative filtering algorithm, this thesis proposed A Collaborative Filtering Recommendation Algorithm with Time-Adjusting Based on Cloud Model (CTCFR).Lastly based on the new algorithm, the thesis uses authority Movielens data collection to experience the algorithm, using the verification of the multiple experimental programs, proves the efficiency and rationality of the proposed algorithm.