Research On Collaborative Filtering Algorithm Based On Global And Local Similarity Measure

The research of this paper is derived from “Web service recommendation servicebased on heterogeneous network analysis”, which is a National Natural ScienceFoundation of China.Nowadays，with the development of information technology and e-commercerecommendation system gradually become a hot topic of e-commerce technology.Among them, the collaborative filtering technology is the most successfulrecommendation technology. From the perspective of enhancing the quality ofrecommendations, this paper makes a research and improvement on similarity measureof collaborative filtering algorithm.Through research, we find that the traditional similarity measures only consider theuser or items directly related, there will be a problem of low accuracy. This paperdivided the similarity into two aspects of global and local, each was designed algorithmto calculate the similarity. The proposed method can reflect the similarity between usersor items more accurate.The details of research works in this paper include:①Analyzed the principle and related technologies of recommendation system,focusing on the technologies of collaborative filtering and discuss the unsolvedproblems in the recommendation system. Then describes the necessity of improving therecommendation algorithm and explains the meaning of collaborative filteringalgorithms by improving the similar measure algorithm.②According to the shortcoming of the low accuracy rate of similarity algorithmsfor collaborative filtering, proposed a way to optimize similarity measure algorithmfrom the global and local information of scoring matrix.③Proposed a similarity measure based on global and local information, throughimproving the Jaccard algorithms and user score preference algorithm to calculate theglobal similarity and calculate the local similarity from three aspects of approximate,similarity and agreement. Finally, integrate the global and local similarity measure asthe proposed GASA algorithm.④Validated the effectiveness of the collaborative filtering recommendationalgorithm based on GASA through the experiment.⑤Designed a recommendation engine of collaborative filtering algorithm based on global and local similarity measure, and implemented a prototype log managementsystem for collecting and analyzing user behavior data.