Research On Recommendation Algorithm Based On Collaborative Filtering

The recommendation system plays a huge role as a link between users and information.Collaborative filtering can be said to be one of the most widely used and most successful algorithms in the recommendation algorithm.Therefore,the focus of this paper is on the collaborative filtering algorithm,which mainly solves two major problems faced by collaborative filtering:data sparseness and low recommendation accuracy.An improved method for the traditional collaborative filtering algorithm and an improved collaborative filtering algorithm based on the restricted Boltzmann machine are proposed.The main work is as follows:1?In order to improve the recommendation accuracy of the current collaborative filtering algorithm,a new similarity calculation method and the influence of the scoring prediction method on the model are proposed in the memory-based collaborative filtering,those are,the mean square error(MSD)similarity calculation and four kinds of Different scoring methods were used,and it was verified in the experiment that the accuracy of the model recommendation was improved.2?Matrix decomposition can effectively solve the problem of data sparseness.Therefore,in this paper,the traditional collaborative filtering algorithm based on singular value decomposition is analyzed,and collaborative least filtering based on cross-least squares(ALS)is also introduced.And use Spark distributed platform to carry out experiments,consider the influence of various parameters of the model on the model effect,and the comparison of the algorithm.3?Combining the deep learning with the recommendation system,this paper analyzes a variety of collaborative filtering algorithms based on the restricted Boltzmann machine.Based on this analysis,an improved fusion real-value RBM model is proposed,which is based on the Python language.The Tensorflow framework is validated.The experimental results show that the prediction accuracy is improved to some extent.