Research And Application Of Intelligent Recommendation System Based On Clustering

In recent years,the Internet industry is constantly expanding and large quantities of data are constantly being produced by the big machine on the Internet.It is quite difficult for users to gather a large amount of valuable and valuable information in mass data heaps.It is very difficult to extract valuable information from huge amounts of data using the general way of collecting information.Recommended system is one of the technical means to solve the above problems,but also the most recent research hotspot.The popular recommendation algorithms are content based recommendation,model based recommendation,collaborative filtering and so on.In order to solve the cold start,sparsity and real-time problems of the currently applied recommender system,this dissertation chooses the cluster-based collaborative filtering algorithm as the main research object.And the shuffled frog leaping algorithm is merged into an improved collaborative filtering algorithm.The specific work of this dissertation is as follows:1.Analyze several classical clustering algorithms.Spectral clustering has the advantages of easy implementation,superior performance,and the ability to identify all shapes of spatial data.Therefore,spectral clustering was selected as the main research clustering algorithm.Based on standard spectral clustering,this dissertation proposes a spectral clustering of maximum distance product.Improved the initial cluster center instability of spectral clustering.The data pre-processing of the previous period was specifically done for the recommendation system,and the user interest preference and matrix pre-filling were added.A pre-populated improved spectral clustering algorithm that fuses user preferences is formed.In the last chapter of this dissertation,we use the UCI standard data set to analyze the improved clustering algorithm.2.The traditional shuffled frog leaping algorithm is analyzed.Aiming at the problem that the convergence speed of the algorithm is not fast enough and it is easy to premature convergence,an improved algorithm of shuffled shuffled frog leaping with normal distribution variation is proposed.The steps of the algorithm to solve the nearest neighbor problem are introduced in detail.For the following and collaborative filtering to lay the foundation.Combined with the improved spectral clustering algorithm,improved frog leaping algorithm,and collaborative filtering,the main recommendation algorithm of this dissertation is given.In the last chapter of this dissertation,theimproved shuffled frog leaping algorithm and recommended algorithm are analyzed by using the standard function and MovieLens data set respectively,and the effectiveness of the algorithm is verified.3.Build a simple personalized movie recommendation platform.The improved recommendation algorithm of this dissertation is used as the core recommendation module,and the actual data is used to test the effect of the algorithm.