Research And Application Of Collaborative Filtering Algorithm Based On Time Decay And Dimensionality Reduction Optimization

Collaborative filtering algorithm is one of the more widely used algorithms in recommendation algorithm research.However,with the increasing amount of data,collaborative filtering algorithms also face many tests.In this paper,we address the influence of time factor on item rating and the sparsity of item-rating matrix.The research proposes a collaborative filtering algorithm that incorporates the time decay function and interference factor and a collaborative filtering algorithm based on principal component analysis and implicit Dirichlet distribution,and finally designs and develops a movie recommendation system based on the above improved algorithms.The main research contents and results are as follows:1.A collaborative filtering algorithm that fuses the time decay function and the interference factor is proposed.The algorithm combines the time decay function and interference theory to temporally weight the item ratings when calculating the item similarity,which solves the influence of time on item ratings and makes the predicted ratings of items more accurate.The case analysis and experimental results together show the feasibility of the collaborative filtering algorithm that combines the time decay function and the interference factor,and reduces the error in predicting item scores and improves the accuracy of recommendations.2.A collaborative filtering algorithm based on principal component analysis and implicit Dirichlet distribution is proposed.First,principal component analysis is used to reduce the dimensionality of the item-rating matrix,which solves the sparsity problem of the rating matrix.Secondly,the implicit Dirichlet distribution is used to generate topics for the item-label matrix and calculate the similarity of items together with the reduced-dimensional scoring matrix.The experimental results show that the collaborative filtering algorithm based on principal component analysis and implicit Dirichlet distribution is effective and feasible,and the accuracy of its recommendation results is improved compared with other collaborative filtering algorithms.3.A movie recommendation system is designed and developed using Spring Boot+My Batis+Vue framework.The system implements the functions of algorithm selection,parameter setting,recommended movies,movie review/rating,information management,etc.The algorithm studied in this paper is used for actual movie recommendation.The research contributions of this paper: Integrate the time decay function and interference theory to propose a new algorithm to accurately predict item ratings and improve recommendation accuracy;improve the calculation of item similarity by using principal component analysis to reduce the dimensionality of the rating sparse matrix,and then generate different topics for the item-label matrix using the implied Dirichlet distribution;and design and implement a movie recommendation system based on the above improved algorithm.