Research And Improvement Of Collaborative Filtering Algorithm Based On Maximum Clique

Due to the rapid development of computer science and technology in the recent decades,the Internet has been used more and more widely and popularized among the masses.However,Internet users to bring more choice and better products at the same time,but also makes the network data show a geometric multiples of growth,leading to the emergence of information overload.Due to the "information overload" phenomenon,the recommender system came into being.There are many problems with the traditional recommendation system: severe cold start and data sparseness problems will cause the recommendation system to be less efficient.The group referral system only considers the influence of the group o n the individual and then makes recommendations,which will cause a new cold start problem for some members who lack social relations.The trustworthiness is difficult to measure because of the close standard of socialized friends in the recommendation sys tem based on trust network.Excessive trust and extrovert easily lead to the user's interest in understanding the deviation.In view of the above problems,we draw on the advantages of the traditional recommendation system and improve the traditional collaborative filtering algorithm by combining the knowledge of graph theory and social network to solve the existing problems.The main contents are as follows:(1)We studied three kinds of maximal regularization methods,such as backtracking method,branch-and-bound method and Bron-Kerbosch method,two new ant colony algorithms and their improved maximal regularization algorithm.By comparing the performance of Bron-Kerbosch algorithm and two kinds of ant colony algorithm,an improved ant colony algorithm with the highest time efficiency is selected as the largest group algorithm to lay the foundation for the following algorithm.(2)Studied the relationship characteristics of social networks,hoping to improve the traditional system of cold start by social ne twork relationship.At the same time,the steps to improve the collaborative filtering algorithm based on the largest group are described in detail.By choosing the largest group member in the social network and the collaborative recommendation system,a hybrid maximum group and collaborative filtering algorithm is proposed to generate the group recommended project set;Finally,the fusion of multiple group scores instead of the original recommended score,and the recommended value of the alternative instead of the recommended value together as a recommended set recommended to each member.(3)In order to verify the effectiveness of this method,we use YELP data to compare two groups of experiments.Through a large number of experimental comparison,we can verify that the improved algorithm based on the largest group of collaborative filtering proposed in this paper has a certain degree of reliability and effectiveness.Compared with the traditional collaborative filtering algorithm,the recommended accuracy rate is about 1%.But also reduced 0.13% of the more serious cold start problems in the traditional recommendation system.