Recommender System Study Based On Log-Likelihood Similarity

With the rapid development of Internet, Information overload leads people to spend a lot of time and vigor in finding fulfilling information. Recommender system can analyze user’s interest automatically and find information which meets the user’s interest in the large database for the user as reference, reduce time and vigor that users spent in finding information. Thus, recommender system has been widely used and plays a huge role in E-commerce, Internet service and social networks. How to improve recommendation’s quality and promote user’s satisfaction is the focus in the research of recommender system.Based on the analysis in recommender system and there kinds of commonly used recommender algorithms, this paper research the theory and implementation method of an LogLikelihood-Based Slope One Improved Algorithm, and use it in a prototype design of a movie recommendation system in order to promote the accuracy rate of predication, and the quality of movie recommendation. The specific research works are as follows:Through the study of the three recommender algorithms commonly used in recommender system, especially the Slope One recommender algorithm and similarity calculating methods, this paper severally compared and analyzed their merit and demerit.Aimed at the deficiency existing in the above algorithm, this paper presented an LogLikelihood-Based Slope One Improved Algorithm -LBSO recommender algorithm. It introduced contingency table to analyze relationship between any two items, calculated the log-likelihood ratio statistic, then converted the log-likelihood ratio statistic to log-likelihood similarity, as the weight of rating prediction formula. Besides, it used an improved neighbor selecting algorithm in combination to exclude dissimilar items for the target item in rating prediction, improving the accuracy rate of target item’s rating prediction. Through the Python programming language to implement the above recommender algorithm.Using the MovieLens dataset to analyze and compare. The experimental results show that the improved algorithm LBSO has a better accuracy rate in rating prediction.Based on the investigation and requirement analysis of movie recommender system, this paper implements a prototype design of movie recommendation system, including architectural structure, functional module and database design etc. Furthermore, this paper adopt MVC model design as system structure, and uses LBSO recommender algorithm to predict movie’s rating which the active user has not yet given. Then LBSO algorithm will sort the candidate movie form high rating to low rating and show a movie recommender list to the user, providing a reasonable basis to make a choice and improving the efficiency of searching movies for users.