java - 我需要一些关于这些公式实现的帮助

Question

我需要一些帮助来实现这些公式。我想我正确地实现了它们，但由于某种原因我没有得到预期的结果:

这分别是NMI、I、H函数的代码。 公式是否正确执行？ 谢谢

int totalN = getTotalN(events);
        double h1 = H(clusters, totalN);
        double h2 = H(events, totalN);
        double valueI = I(clusters, events, totalN);
        double value_NMI = valueI / (double) ((h1 + h2) / (double) 2);
        System.out.println("NMI: " + value_NMI);

static public double I(HashMap<String, ArrayList<String>> clusters, HashMap<String, ArrayList<String>> events, int totalN) {

        //store sorted content to contents
        Iterator<Map.Entry<String, ArrayList<String>>> it = events.entrySet().iterator();
        Iterator<Map.Entry<String, ArrayList<String>>> it2 = clusters.entrySet().iterator();

        String key;
        ArrayList<String> event;
        ArrayList<String> cluster;

        double valueI = 0;
        while (it.hasNext()) {

            Map.Entry<String, ArrayList<String>> mapItem = it.next();
            key = mapItem.getKey();

            //if cluster doesn't exist
            //if(!clusters.containsKey(key)) continue;
            //cluster = clusters.get(key);

            event = mapItem.getValue();

            while (it2.hasNext()) {
                Map.Entry<String, ArrayList<String>> mapItem2 = it2.next();
                cluster = mapItem2.getValue();

            float common_docs = 0;
            for (int i=0; i< event.size(); i++) {   
                for (int j=0; j< cluster.size(); j++) { 
                    if (event.get(i).equals(cluster.get(j))) {
                        common_docs = common_docs + 1;
                        break;
                    }
                }
            }


            if (common_docs != 0) valueI = valueI + ( ( common_docs / (float) totalN) * Math.log((common_docs * totalN) / (float) (event.size() * cluster.size())) );       
            }
        }

        return valueI;
    }


    static public double H(HashMap<String, ArrayList<String>> clusters, int totalN) {

        //store sorted content to contents
        Iterator<Map.Entry<String, ArrayList<String>>> it = clusters.entrySet().iterator();
        ArrayList<String> cluster;

        double entropy = 0;
        while (it.hasNext()) {

            Map.Entry<String, ArrayList<String>> mapItem = it.next();
            cluster = mapItem.getValue();

            double ratio = cluster.size() / (float) totalN;
            entropy = entropy + ratio * Math.log(ratio);

        }

        return -entropy;
    }

    static public int getTotalN(HashMap<String, ArrayList<String>> dataset) {

        int totalN = 0;
        Iterator<Map.Entry<String, ArrayList<String>>> it = dataset.entrySet().iterator();
        ArrayList<String> item;

        while (it.hasNext()) {

            Map.Entry<String, ArrayList<String>> mapItem = it.next();
            item = mapItem.getValue();

            for (int i=0; i< item.size(); i++) {
                totalN = totalN + 1;
            }

        }

        return totalN ;
    }

Answer 1

我猜没有。我刚刚检查了 I(C,E)，并且您不会在它的每次迭代中都重置 it2，这对于嵌套总和是必需的。