Research And Application Of Density-Based Clustering Algorithms

With the development of information technology and the convenience ofinformation acquisition，we are overwhelmed by the amount of information.It is anurgent problem need to be solved that how to extract knowledge of interest andcomplete a specific task from so much knowledge.Based on this demand,Data Miningtechnology that helps people to analyze the amount of knowledge which impliesvaluable models and technology comes into being.Clustering analysis is a veryimportant technology in data mining.There are both theoretical significance and value ofapplication in study of it deeply.With the application in Enterprise Competitive Intelligence System,this paper madethe following improvements for the problem of DBSCAN algorithms.(1)During the extraction of the seed objects,if the seed object is not core object,butthe number of objects in it’s neighborhood region is close to MinPts,we can classify thisseed object into this cluster. By the experiment,we set this value as0.85MinPts.(2)It doesn’t cause objects’ loss when we are selecting seed object in core object’sregion.But when we extract the seed objects,there will be loss.So,we choose the seedobject far from the core object.In this paper we choose the seed object between0.95εand ε from the core object.In the last,on the base of theory studied above and combining with the project，weapply this Improved algorithms to Enterprise Competitive Intelligence System andachieved good results.