Parallel Collaborative Filtering Recommendation Algorithm Based On Social Networks And Key Users

In the era of big data,it is becoming more and more important to extract key information efficiently,and the recommendation system is one of the important means to effectively extract information from massive data.According to the user's personal information,historical behavior and other data,it can accurately judge the user's interests and hobbies then provide personalized recommendation services for users.Nowadays,the research on recommendation system is in full swing,especially the collaborative filtering recommendation algorithm,which has become one of the most common and effective recommendation algorithms.However,there are also many problems in the collaborative filtering recommendation algorithm,such as cold start,data sparseness,time extension,and unsatisfactory recommendation effect.Therefore,in view of the above problems,this paper proposes a parallel collaborative filtering recommendation algorithm that integrates social networks with key users,and improves the performance of collaborative filtering recommendation algorithms from three aspects.Converging social networking and key user solutions for cold start and data sparse issues.The social network trust matrix,key user project scoring matrix and user project scoring matrix are obtained by preprocessing the data set,and by using multi-source data fusion to effectively solve the problems that item cold start?user cold start and alleviate data sparseness in the recommended system.Considering the delay of massive data processing,a parallel computing strategy based on Spark distributed cluster is proposed,and the parallelization of the recommended algorithm is realized,which can effectively shorten the running time and improve the user experience.According to the problem of unsatisfactory recommendation effect,the similarity calculation method is improved,and the modified cosine similarity calculated by social network data,key user score data and user item score data is weighted and fused to obtain more accurate and stable similarity calculation formula.It can be seen from the experimental results that when the accuracy of the similarity calculation is improved,the recommended results are more in line with the user's interests.The experiment uses Epinions public dataset.The results of multiple sets of comparison experiments show that the proposed collaborative collaborative filtering recommendation algorithm for social network and key users can effectively improve recommendation accuracy,shorten running time and optimize recommendation effect.This paper has 18 figures,9 tables and 71 references...