Data Mining Clustering Algorithm--dbscan

The data set looks like this:

1,1,12,1.5,13,0.5,13,5,-17,0.75,-17,4,28,5,28,5.5,2

The dataset has three properties, namely X and y in two-dimensional coordinates, the third attribute is the owning class, and 1 is the outlier, and the coordinate system looks like this:

The source code is as follows:

 PackageNeugle.dbscan;ImportJava.io.BufferedReader;ImportJava.io.FileReader;Importjava.util.ArrayList;Importjava.util.List;ImportJava.util.Random; Public classDBScan {PrivateList<point> pointlist =NewArraylist<dbscan.point> ();//Sample Data read- in    PrivateList<list<point>> clusterlist =NewArraylist<list<point>> ();//Final Classification Results    PrivateList<point> noiselist =NewArraylist<dbscan.point> ();//Noise Data collection    PrivateList<point> npointlist =NewArraylist<dbscan.point> ();//Candidate Data Collection    PrivateList<integer> unvisitedlist =NewArraylist<integer> ();//unvisited Collection    Private DoubleEPs//Neighborhood Radius    Private intminpts;//density    classPoint { Public Doublex;  Public Doubley;  PublicString Point_type;  Public Booleanisvisited =false; }     PublicDBScan (DoubleEpsintminpts) {         This. EPS =EPS;  This. minpts =minpts; }    //reading Data     PublicList<point>ReadFile (String filePath) {FileReader fr=NULL; BufferedReader BR=NULL; Try{FR=NewFileReader (FilePath); BR=NewBufferedReader (FR); String Line=NULL;  while(line = Br.readline ())! =NULL) { point point=NewPoint (); String[] AGRs= Line.split (","); Point.x= Double.parsedouble (agrs[0]); Point.y= Double.parsedouble (agrs[1]); Point.point_type= Agrs[2];  This. Pointlist.add (point); }        } Catch(Exception e) {e.printstacktrace (); } finally {            Try{br.close (); } Catch(Exception e) {e.printstacktrace (); }        }        returnpointlist; }    //Dbscan Main Method     Public voidDbscanfun (String filePath) { This.        ReadFile (FilePath); //This . Norm ();         while( This. Isover ()) {point P= This. Pointlist.get ( This. Randomnum ());//randomly find the unvisited nodep.isvisited =true;//Mark P for visitedList<point> neighborlist = This. Getneighbors (P);//Find a neighbor node that meets the minimum density            if(Neighborlist.size () < This. minpts) {//Adding noise data                 This. Noiselist.add (P);//put p into the noise collection}Else{List<Point> CList =NewArraylist<dbscan.point> ();//Create a new cluster CClist.add (P);//put P in cluster C                 This. npointlist = neighborlist;//A collection of neighborhood objects that make N p                 for(inti = 0; I < This. Npointlist.size (); i++) {                    if( This. Npointlist.get (i). isvisited = =false) {//Find the unvisited node in P '                         This. Npointlist.get (i). isvisited =true;//Mark P ' for visitedList<point> neighborlists = This                                . Getneighbors ( This. Npointlist.get (i));//calculates P ' satisfies the neighbor's node collection                        if(Neighborlists.size () >= This. minpts) {                             for(intj = 0; J < Neighborlists.size (); J + +) {                                 This. Npointlist.add (Neighborlists.get (j));//Add the Neighborhood node of P ' to n}} clist.add ( This. Npointlist.get (i));//add P ' to cluster C                    }                }                 This. Clusterlist.add (CList); }        }    }    //randomly selected in an inaccessible collection    Private intRandomnum () {intnum = This. Unvisitedlist.size (); Random Rand=NewRandom (); intRandnum =rand.nextint (num); return  This. Unvisitedlist.get (Randnum); }    //get a neighborhood collection    PrivateList<point>getneighbors (point P) {List<Point> list =NewArraylist<dbscan.point>();  for(inti = 0; I < This. Pointlist.size (); i++) {            DoubleValue = This. Distancecalculate ( This. Pointlist.get (i), p); if(Value! = 0 && Value <= This. eps) {List.add ( This. Pointlist.get (i)); }        }        returnlist; }    //Euclidean distance formula    Private Doubledistancecalculate (Point iris1, point Iris2) {Doublesum = math.sqrt (Math.pow (iris1.x-iris2.x), 2)                + Math.pow ((IRIS1.Y-IRIS2.Y), 2)); returnsum; }    //determine if the data has been accessed    Private BooleanIsover () { This. unvisitedlist =NewArraylist<integer>();  for(inti = 0; I < This. Pointlist.size (); i++) {            if( This. Pointlist.get (i). isvisited = =false) {unvisitedlist.add (i); }        }        if( This. Unvisitedlist.size () > 0) {            return true; }        return false; }     Public voidPrint () {System.out.println ("Poly to" + This. Clusterlist.size () + "class");  for(inti = 0; I < This. Clusterlist.size (); i++) {List<Point> C = This. Clusterlist.get (i); System.out.println ("------------");  for(intj = 0; J < C.size (); J + +) {System.out.println (C.get (j). x+ "" + C.get (j). Y + "" +C.get (j). Point_type);            } System.out.println (C.size ()); System.out.println ("------------"); } System.out.println ("Noise Point has" + This. Noiselist.size () + "X"); System.out.println ("------------");  for(inti = 0; I < This. Noiselist.size (); i++) {System.out.println ( This. Noiselist.get (i). x + "" + This. Noiselist.get (i). Y + "" + This. Noiselist.get (i). Point_type); } System.out.println ("------------"); }     Public Static voidMain (string[] args) {DBScan C=NewDBScan (2.5, 2); C.dbscanfun ("D:\\data\\dbscan\\test.data");    C.print (); }}

The experimental results are as follows:

Poly 2 class------------8.0 5.5 27.0 4.0 28.0 5.0------------------------3.0 0.5 11.0 1.0 12.0 1.5------------Noise point has 2-------- ----3.0 5.0-17.0 0.75-1------------

　　

Data Mining Clustering Algorithm--dbscan