Research On Collaborative Filtering Recommender Algorithm Based On Ontology And Dimensionality Reduction Technology

With the expansion of the scale of information systems and the rapid development of e-commerce,the number of users and item ratings has increased dramatically,the scale of user ratings matrix has increased dramatically,how to deal with high-dimensional matrices,and to deal with the problem of data sparsity,enhance the scalability of the recommender system and the real-time performance of the recommender algorithm have become the focus of current research.This research combines the two technologies of ontology and dimensionality reduction to try to solve the above problems.The research proposes a collaborative filtering(CF)recommend er algorithm based on ontology and dimensionality reduction technology.Taking movies as an example,the Movie O ntology(MO)is used to calculate the semantic similarity of items based on domain ontology,and the similarity of items based on singular value decomposition(SVD)rating matrix is integrated to determine the weighting factors of the two to generate item-based predictions rating;re-fusion based on the user's predictions rating,determine the fusion factor,generate the final predictions rating,and generate recommenders.This algorithm is different from the traditional CF algorithm that only considers the user's rating of the rating matrix,introduces the semantic similarity of the item based on the ontology,considers the impact of the item's own attributes on the user's rating,so as to reconcile the item similarity and reduce the sparseness of the rating matrix data impact,improve the accuracy of recommendations.At the same time,for the scalability of the traditional CF algorithm,the Expectation Maximization(EM)algorithm is used to cluster user-item ratings,and similar users and items are divided into clusters.The similarity of objects between the same clusters is similarity high,and the similarity of objects between different clusters is low.It is not necessary to traverse all the objects in the data set when obtaining the nearest neighbor.Thereby greatly reducing the calculation time,improving real-time performance and enhancing the scalability of the recommender algorithm.Finally,the CF recommender algorithm based on ontology and dimensionality reduction technology is verified by Movie Le ns datasets.Evaluation indicators include: Throughput,Mean Absolute Error(MAE),Precision,Recall,and F1 value.Using Matlab to realize the calculation and recommendation of the algorithm,and setting up a comparison experiment group,comparing the experimental results with the traditional algorithm,it is proved that the CF recommender algorithm based on ontology and dimensionality reduction technology effectively improves the data sparsity and scalability issues.And improve the recommender accuracy.