La classe CLUSTER
La classe CLUSTER
Embedded Files
Un cluster contient
Un centroïde de type Data qui représente la donnée centrale,
Un ensemble de données
Un numéro.
On peut ajouter des données statistiques telles que : la distance minimum d’une donnée au centroïde, la distance maximale, la distance moyenne, ...
1. class Cluster {
2. /**data associtaed to the cluster*/
3. ArrayList<Data> dataSet;
4. /**computed data that reprensent the center of the cluster*/
5. Data centroid;
6. /**nb of clusters*/
7. private static int nb;
8. /**no of the cluster*/
9. int no;
10. /**minimal distance between one of the data and the centroid */
11. double minDist;
12. /**maximal distance between one of the data and the centroid */
13. double maxDist;
14. /**average of the distances between the data and the centroid */
15. double moyDist;
16.
17. public Cluster(){
18. dataSet = new ArrayList<>();
19. centroid = new Data();
20. no = nb++;
21. }
22.
23. /**initialise the data with a centroid*/
24. public Cluster(Data _centroid) {
25. this();
26. centroid = _centroid;
27. }
28.
29. /**add a data to the cluster (and tell to the data that it is associated to this cluster)*/
30. public void add(Data data) {
31. dataSet.add(data);
32. data.setCluster(this);
33. }
34.
35. /**remove a data to from the cluster
36. * (and tell to the data that it is no more associated to this cluster)*/
37. public void remove(Data data) {
38. dataSet.remove(data);
39. data.setCluster(null);
40. }
41. }
La fonction principale de la classe est celle recalculant les coordonnées du centroide afin que chacune représente la moyenne des coordonnées des données du cluster.
1. /**recompute the center of the cluster*/
2. public void centralize() {
3. int nbElt = dataSet.size();
4. if(nbElt>0) {
5. int dim = dataSet.get(0).length;
6. //we can't pass a non final value to an inner class (like with lambda expression), so we pass by a small array
7. int[] tI = {0};
8. for(int i=0; i<dim;i++) {
9. tI[0] = i;
10. //sum of normalized ith values
11. double sumI = dataSet.stream().mapToDouble(d->d.getNormValue(tI[0])).sum();
12. double average = sumI/(double)nbElt;
13. if(centroid.getNormValue(i)!=average) {
14. centroid.setNormValue(i, sum/(double)nbElt);
15. //sum of ith values
16. sumI = dataSet.stream().mapToDouble(d->d.getValue(tI[0])).sum();
17. centroid.setValue(i, sumI/(double)nbElt);
18. }
19. }
20. }
21. }
Ajoutons une fonction pour le calcul des stats et pour la conversion du cluster en chaîne de caractères :
1. /**compute the stats (minimal distance from a data to the centroid, maximal distance, average of the distances)*/
2. public void computeStats() {
3. double somDist=0;
4. minDist=Double.POSITIVE_INFINITY;
5. maxDist=Double.NEGATIVE_INFINITY;
6. for(Data data:dataSet) {
7. double dist = data.distNorm(centroid);
8. if(dist<minDist) minDist=dist;
9. if(dist>maxDist) maxDist=dist;
10. somDist+=dist;
11. }
12. if(dataSet.size()>0) moyDist = somDist/dataSet.size();
13. }
14. /**@return the no of the clusters, its nb of data, the stats and the data*/
15. @Override
16. public String toString() {
17. StringBuilder sb = new StringBuilder("cluster " + no + ", nb elts = " + dataSet.size() + "\n");
18. sb.append("--> centroid = ").append(centroid).append("\n data");
19. if (dataSet.size()<50)
20. for(Data data:dataSet) sb.append(data.toString()).append("\n");
21. sb.append("--> dist min=").append(String.format(Locale.ENGLISH,"%.2f", minDist));
22. sb.append("; dist max=").append(String.format(Locale.ENGLISH,"%.2f", maxDist));
23. sb.append("; average dist=").append(String.format(Locale.ENGLISH,"%.2f", moyDist));
24. sb.append(" \n---- ");
25. return sb.toString();
26. }
Page updated
Google Sites
Report abuse