Collaborative Filtering Recommendation Algorithm Based On Granular Layer Clustering

User-based collaborative filtering recommendation collects and analyzes historical data related to user behavior,and implements user recommendation based on the similarity with other users.However the user’s historical behavior data set has serious sparseness,which reduces the accuracy of the recommendation.K-means clustering and fuzzy C-means clustering reduce the impact of sparse data,but the recommended diversity effect is not obvious enough,integrating granular clustering into collaborative filtering recommendation can not only solve the algorithm’s sensitivity to sparse data,but also realize diversity recommendation.Granular computing is a calculation tool that simplifies complex problems.Through the selection of appropriate granularities,the optimal solution of the original problem can be obtained by solving different granular layers.The similarity of users is used as the threshold of granular layer division.By adjusting the threshold,the appropriate granularity can be selected to achieve the global optimum,so as to improve the diversity and accuracy of recommendation.The specific contents are as follows:First of all,the experiment reproduced five algorithms based on user attributes and model clustering,through data comparison,the sensitivity of each algorithm to data sparseness and the sensitivity of data set size is analyzed.For large data sets,these five algorithms reduce the recommendation effect.Second,the data preprocessing technology is used to reduce the data sparsity of collaborative filtering recommendation,and the SMOTE technology is used to fill in blank data,which improves the accuracy of recommendation.At the same time,we explored the influence of the time factor on the recommendation quality,and found that whether the time factor affects the recommendation effect depends on the selection process of the cluster center.Third,aiming at the poor diversity problem faced by the recommendation system,a covering rough granular layer clustering model is proposed.This model reduces the impact of data sparseness.It finds the user’s local rough granular set and then obtains the global covering rough granular set.This model reduces the local optimal problem caused by data sparseness,and can provide users with accurate multi-level and multi-granularity recommendations,thereby realizing the diversity of recommendations.The application of granular computing theory to the recommendation algorithm is an optimization and innovation of the recommendation model.Through the experimental test of the benchmark data set,it is proved that for large data sets,the recommendation model based on granular computing can achieve the diversity of recommendations without reducing the accuracy of recommendations.Figure 10;Table 32;Reference 67...