我正在使用k-means算法进行数据聚类和使用大数据集。我有近10万篇研究论文，我想通过使用k-means来聚类它们。我使用传统的k-means，也使用倒排索引，但在两个程序中。当我放入50000时，我在Java中使用NetBeans（使用NetBeans）。
代码：
import java.util.*;
import java.io.*;
public class kmeans
{
public static void main(String[]args) throws IOException
{
//read in documents from all .txt files in same folder as kmeans.java
//save parallel lists of documents (String[]) and their filenames, and create global set list of words
ArrayList<String[]> docs = new ArrayList<String[]>();
ArrayList<String> filenames = new ArrayList<String>();
ArrayList<String> global = new ArrayList<String>();
File folder = new File(".");
List<File> files = Arrays.asList(folder.listFiles(new FileFilter() {
public boolean accept(File f) {
return f.isFile() && f.getName().endsWith(".txt");
}
}));
BufferedReader in = null;
for(File f:files){
in = new BufferedReader(new FileReader(f));
StringBuffer sb = new StringBuffer();
String s = null;
while((s = in.readLine()) != null){
sb.append(s);
}
//input cleaning regex
String[] d = sb.toString().replaceAll("[\\W&&[^\\s]]","").split("\\W+");
for(String u:d)
if(!global.contains(u))
global.add(u);
docs.add(d);
filenames.add(f.getName());
}
//
//compute tf-idf and create document vectors (double[])
ArrayList<double[]> vecspace = new ArrayList<double[]>();
for(String[] s:docs){
double[] d = new double[global.size()];
for(int i=0;i<global.size();i++)
d[i] = tf(s,global.get(i)) * idf(docs,global.get(i));
vecspace.add(d);
}
//iterate k-means
HashMap<double[],TreeSet<Integer>> clusters = new HashMap<double[],TreeSet<Integer>>();
HashMap<double[],TreeSet<Integer>> step = new HashMap<double[],TreeSet<Integer>>();
HashSet<Integer> rand = new HashSet<Integer>();
TreeMap<Double,HashMap<double[],TreeSet<Integer>>> errorsums = new TreeMap<Double,HashMap<double[],TreeSet<Integer>>>();
int k = 3;
int maxiter = 500;
for(int init=0;init<100;init++){
clusters.clear();
step.clear();
rand.clear();
//randomly initialize cluster centers
while(rand.size()< k)
rand.add((int)(Math.random()*vecspace.size()));
for(int r:rand){
double[] temp = new double[vecspace.get(r).length];
System.arraycopy(vecspace.get(r),0,temp,0,temp.length);
step.put(temp,new TreeSet<Integer>());
}
boolean go = true;
int iter = 0;
while(go){
clusters = new HashMap<double[],TreeSet<Integer>>(step);
//cluster assignment step
for(int i=0;i<vecspace.size();i++){
double[] cent = null;
double sim = 0;
for(double[] c:clusters.keySet()){
double csim = cosSim(vecspace.get(i),c);
if(csim > sim){
sim = csim;
cent = c;
}
}
clusters.get(cent).add(i);
}
//centroid update step
step.clear();
for(double[] cent:clusters.keySet()){
double[] updatec = new double[cent.length];
for(int d:clusters.get(cent)){
double[] doc = vecspace.get(d);
for(int i=0;i<updatec.length;i++)
updatec[i]+=doc[i];
}
for(int i=0;i<updatec.length;i++)
updatec[i]/=clusters.get(cent).size();
step.put(updatec,new TreeSet<Integer>());
}
//check break conditions
String oldcent="", newcent="";
for(double[] x:clusters.keySet())
oldcent+=Arrays.toString(x);
for(double[] x:step.keySet())
newcent+=Arrays.toString(x);
if(oldcent.equals(newcent)) go = false;
if(++iter >= maxiter) go = false;
}
System.out.println(clusters.toString().replaceAll("\\[[\\w@]+=",""));
if(iter<maxiter)
System.out.println("Converged in "+iter+" steps.");
else System.out.println("Stopped after "+maxiter+" iterations.");
System.out.println("");
//calculate similarity sum and map it to the clustering
double sumsim = 0;
for(double[] c:clusters.keySet()){
TreeSet<Integer> cl = clusters.get(c);
for(int vi:cl){
sumsim+=cosSim(c,vecspace.get(vi));
}
}
errorsums.put(sumsim,new HashMap<double[],TreeSet<Integer>>(clusters));
}
//pick the clustering with the maximum similarity sum and print the filenames and indices
System.out.println("Best Convergence:");
System.out.println(errorsums.get(errorsums.lastKey()).toString().replaceAll("\\[[\\w@]+=",""));
System.out.print("{");
for(double[] cent:errorsums.get(errorsums.lastKey()).keySet()){
System.out.print("[");
for(int pts:errorsums.get(errorsums.lastKey()).get(cent)){
System.out.print(filenames.get(pts).substring(0,filenames.get(pts).lastIndexOf(".txt"))+", ");
}
System.out.print("\b\b], ");
}
System.out.println("\b\b}");
}
static double cosSim(double[] a, double[] b){
double dotp=0, maga=0, magb=0;
for(int i=0;i<a.length;i++){
dotp+=a[i]*b[i];
maga+=Math.pow(a[i],2);
magb+=Math.pow(b[i],2);
}
maga = Math.sqrt(maga);
magb = Math.sqrt(magb);
double d = dotp / (maga * magb);
return d==Double.NaN?0:d;
}
static double tf(String[] doc, String term){
double n = 0;
for(String s:doc)
if(s.equalsIgnoreCase(term))
n++;
return n/doc.length;
}
static double idf(ArrayList<String[]> docs, String term){
double n = 0;
for(String[] x:docs)
for(String s:x)
if(s.equalsIgnoreCase(term)){
n++;
break;
}
return Math.log(docs.size()/n);
}
}
答案 0 :(得分：0)
您可以使用-Xmx ...作为java.exe的命令行参数来增加堆内存量;例如-Xmx1024m将最大堆大小设置为1024Mb。
答案 1 :(得分：0)
您宁愿使用Lucene索引并使用TermFreqs和TermFreqVectors并从索引计算余弦相似度。这将减少您的处理时间和内存问题。你应该试一试。