Research On Collaborative Filtering Algorithm Based On Pareto Optimal And Trust Relationship

Collaborative Filtering(CF)is one of the widely used recommended methods and has been used in many areas.The method is based on the assumption that similar users have similar tastes and interests.Therefore,the CF uses the preferences of users who have similar tastes to the recommended users to obtain useful recommendations.To this end,the historical data information given by the user on the project is used to first find a similar user and then make a prediction.There are two main problems with CF:data sparsity and cold-start issues.Compared to all possible scores,data sparsity is caused by too few projects that users participate in.A cold start issue is an item(or user)that does not have enough previous rating history.Under a cold start project(or user),the system usually does not provide high quality advice.In order to alleviate the above two problems,the main research contents of this paper are as follows:[Objective]To obtain the most similar users of users in the recommendation process by using Pareto optimal concept and trust relationship.It can improve the calculation accuracy of user similarity in traditional collaborative filtering recommendation,thereby optimizing the clustering effect of similar users,and let some business platforms produce better recommendation results,which can alleviate the data sparse and cold start problems that often occur in collaborative algorithms.The interference caused.[Research content]This article considers the Pareto optimal thinking in economics and the trust relationship in social relations into the collaborative filtering recommendation algorithm.The method steps are as follows:First,the trust network is determined according to the trust relationship,and the predicted score set is generated;secondly,according to the generated score set,the adjustment factor is calculated to improve the traditional collaborative filtering algorithm to obtain similar values;and third,the use of Pare The optimal principle identifies the dominant user,reconstructs the trust network,and uses the reconstructed trust network to calculate the similarity value.Fourth,predict the score and implement the recommendation.The addition of adjustment factors and the use of the Pareto concept are the innovations of this paper.[Research method]Firstly,the correlation coefficient is calculated by using Pearson correlation coefficient,then the trust network is reconstructed by Pareto optimal concept.Finally,the experimental evaluation is carried out by the average absolute error index.Through the two experimental cases at the end of the article and the last small empirical analysis,it can be preliminarily concluded that the improved method mentioned in this paper and the traditional collaborative filtering algorithm improve the accuracy and further alleviate the cold.Startup puzzles and sparse puzzles.[Limitations]This paper focuses on the innovation of methods,so the empirical part seems relatively thin,although there are many real data sources,but after screening,the amount of data involved in the experiment is not large,this is the weakness of this article..[Conclusion]The trust relationship between a large number of users contains particularly valuable information.Through the conversion between quality and quantity,the trust information between users is quantified,and the similar values are calculated in combination with other information of the user,and then the Pare is used.The optimal principle is to filter the most trusted users,optimize the traditional Pearson formula,and implement collaborative filtering recommendations.