| The rapid development of the modern science and technology and society,computer has been widely used in public life,especially in recent years,the development of new technology,make people produce and collect data on the ability of rapid growth.At the same time database Zo rate rapidly increase.With the continuous breakthroughs in electronic commerce platform continues to grow,the Internet technology and social networks increasingly popular the boundary and scope of application of Internet has been greatly expanded and consequent estate gave birth to a wide variety of data.Therefore,it is almost impossible to using conventional database analysis technique to find out the valuable content.Effective mining methods are necessary for mining potentially useful information from large data.User demand more and more diverse,can not meet the demand for personalized products will be more a lack of competitiveness,and want to provide such products.First of all,we should to your users,in-depth digging stubborn you user characteristics.This paper is based on the idea that the data mining method is divided into user groups,so as to provide more effective service for users.And K-means is one of the most commonly used data mining algorithms,this paper is to study this method through the study of user segmentation.By the study and analysis of K-means algorithm,the algorithm in the presence of initial clustering center selection way is not reasonable and simple Euclidean distance is not enough to describe the actual problems found.Aiming at the two problems,this paper presents a based on density adaptive grid method to solve the initial clustering center selection problem;according to the Euclidean distance can not reflect the reality of the problem,this paper proposes a kind of improved attribute weighted distance based on representation,the facts show that the improved distance calculation method can better simulate the real world.At last,the research work of this paper is summarized and prospected. |
