Research On Clustering Algorithm Based On Variable Grids

Clustering analysis in data mining is an important data mining technology.In recent years,with the rapid development of the Internet,finance,e-commerce and other industries,cluster analysis has been very extensive research and application.Clustering is the process of dividing a data set into subsets according to some rules.It has been widely studied by many scholars both at home and abroad due to its simple,easy to understand and fast calculation speed.However,the k-means method also has a lot of deficiencies: it is necessary to first determine the number of k;if noise and outliers are selected,they are very sensitive;only convex spherical clusters or very different clusters can be found.Aiming at the shortcomings of k-means clustering,this paper presents a new method of k-means clustering with variable grid optimization.This method can solve the problem that the initial center point is more sensitive to selection by variable grid division of the data set.At the same time,it can effectively find non-convex shape clusters and effectively improve the clustering quality.And the maximum grid density is not the case has been resolved.For k-means algorithm,the number of artificially designated clusters k is insufficient.On the basis of the new k-means clustering method based on variable grid optimization,the optimal clustering number k is obtained by combining the validity indexes.For the lack of dynamic incremental data processing in clustering,a dynamic incremental k-means clustering method based on variable grid is proposed,which has good scalability and high efficiency.For the traditional collaborative filtering recommendation project for the target users need to consider the impact of all users history feedback information on the similarity of the project,leading to the problem of low recommendation quality.The algorithm proposed in this paper compares the features of variable grid clustering with the features of collaborative filtering.Combining the recommendation of commodities,a collaborative filtering recommendation algorithm based on variable grid clustering was proposed,which improved the quality of recommendation system.