Research Of Recommedation Algorithm Based On Clustering Analysis And Collaborative Filtering Technology

With the vigorous development of e-commerce technology,personalized recommendation system has ushered in a good time for rapid development,the research and application of recommendation algorithm has also recevived widespread attention.However,in recent years,network structure is more and more complex,resulting in excessive expansion of network data.In the face of massive network data,how to quickly provide users with accurate personalized recommendation services has become an important topic in the application research of recommendation system.This thesis summarizes the development process,related concepts and theories of the recommended system,and analyzes the commonly used strategies and their advantages and disadvantages.At present,although the collaborative filtering recommendation algorithm has a good application scenario,there are still some shortcomings.Therefor,this thesis presents a number of improvements:First,in order to alleviate the problem that the similarity calculation is not accurate under the sparse data,this thesis presents an improved similarity measure method.The most important step in the collaborative filtering recommendation algorithm is the calculation of the similarity.In the case of data sparse,the results calculated by the traditional similarity measure are often inconsistent with the actual result.To this end,this thesis uses the common preference between users and the rating deviation of common item,the traditional similarity measurement method has been improved;on this basis,an ISCF algorithm based on similarity improvement has been proposed.The experimental results show that the similarity measure method proposed by this thesis can effectively alleviate the inaccurate problem of similarity calculation.Second,for the sparseness of data,this thesis proposes an improved collaborative filtering recommendation algorithm.As the user rating matrix is very sparse,the traditional collaborative filtering recommendation algorithm is difficult to give a satisfactory recommendation results.Therefore,from the point of view of reducing the sparseness of user rating matrix,this thesis uses the clustering analysis and forecasting filling technique to predict and fill the original sparse rating matrix.Then,on the basis of the ISCF algorithm,a KSCF algorithm based on K-means clustering has been proposed.The experimental results show that the algorithm proposed by this thesis can alleviate the influence of the sparseness of data and improve the recommendation quality.In addition,in this thesis,all the experimental methods are offline experiment,and the experimental dataset is MovieLens 100 K dataset.