Research On Model-Based Collaborative Filtering Algorithm In Recommender System

Nowadays,the increasing amount of networkdata and the development of big data technology have brought development opportunities and challenges to recommendation systems.Model-based collaborative filtering has become a mainstream algorithm in recommendation systems.Model-based collaborative filtering algorithms are primarily modeled and solved through machine learning and data-mining model ideas.The principle of the matrix decomposition FunkSVD algorithm is simple and effective,which pushes the model-based collaborative filtering algorithm to a new level.However,when dealing with big-data calculations,data sparseness and iterative oscillations often affect the accuracy of the FunkSVD algorithm.Moreover,when the data volume is in units of GB or more,the FunkSVD algorithm runs slowly and is ineffective.In summary,it is particularly momentous to improve the accuracy and effectiveness of FunkSVD.Improving the accuracy of the FunkSVD algorithm is mainly achieved from two aspects: alleviating data sparseness and iterative shock.The effectiveness of the FunkSVD algorithm is improved mainly through parallel processing of the algorithm.Based on these two parts,this study solves the shortcomings of previous research methods and proposes an improved FunkSVD algorithm,and improved FunkSVD parallel algorithm based on Sparkand GPU.The specific research content is as follows:1.An improved FunkSVD algorithm is proposed to improve the accuracy of the FunkSVD algorithm.The traditional FunkSVD is improved using the deep learning optimization algorithm RMSProp.Considering that the RMSProp algorithm is an improvement of the gradient descent method and is suitable for dealing with sparse data,the RMSProp algorithm is combined with the FunkSVD algorithm to solve the iterative oscillation problem while reducing the impact of data sparseness on the accuracy and provides the global optimal value,thus,the prediction accuracy is improved.2.To solve the problem of slow computation speed of the improved FunkSVD algorithm when processing big data,Spark-based and GPU-based parallel algorithms are proposed.Two big data computing frameworks,Sparkand GPU,can parallelize the process of updating parameters and calculating the inner product of the improved FunkSVD,reducing the computation time of the algorithm and improving the operational efficiency of the algorithm,thus improving the effectiveness of the algorithm.