java - weka java代码kmeans聚类

Question

我使用数据挖掘技术(kmeans 聚类)编写了用于恶意软件检测的 Java 代码。我使用jnetpcap库来嗅探数据包来分析它，然后在方法nextpacket中我为第一个数据包编写了kmeans聚类算法。

该算法运行良好，因为它从类 Instances 创建具有特定属性的对象，这些属性将基于它们进行聚类，但在下一个数据包上，此代码无法再次运行，因为它会引发异常.

我使用的代码是这样的:

for(int dim = 0; dim < numDimensions; dim++)
{
    Attribute current = new Attribute("Attribute" + dim , dim);

    if(dim == 0)
    {
        for(int obj = 0; obj < numInstances; obj++)
        {
            // instances.add(new SparseInstance(numDimensions));
            instances.add(new DenseInstance(numDimensions) );
        }
    }

    for(int obj = 0; obj < numInstances; obj++)
    {
        instances.get(obj).setValue(current, (Double)data[dim+1][obj]);
    }

    atts.add(current);
}

Instances newDataset = new Instances("Dataset" , atts, instances.size());       //this is the line that throws the exception

for(Instance inst : instances)
    newDataset.add(inst);
SimpleKMeans kMeans = new SimpleKMeans();
kMeans.setNumClusters(2);
// kMeans.setMaxIterations(4);
kMeans.buildClusterer(newDataset);
//   int clusterNumbers;
// clusterNumbers=kMeans.numberOfClusters();
for (int j=0;j<numInstances;j++)
{ 
    int classif=kMeans.clusterInstance(newDataset.get(j));
    //  double []distr=kMeans.distributionForInstance(newDataset.firstInstance());
    System.out.println(classif);
    //   System.out.println(distr[0]);
    //  System.out.println(distr[1]);
    ArrayList<Double> temp5=flowFeatures.get((JFlowKey)data[0][j]);
    if (classif==0)
    {
        // instances0.add(newDataset.get(j));
         instance0FlowFeatures.put((JFlowKey)data[0][j], temp5);
    }
    else if(classif==1)
    {
        //instances1.add(newDataset.get(j));
        instance1FlowFeatures.put((JFlowKey)data[0][j], temp5);
    }
}

我看到的异常(exception)是:

java.lang.IllegalArgumentException: Attribute names are not unique! Causes: 'Attribute0' 'Attribute1' 'Attribute2' 'Attribute3' 'Attribute4' 'Attribute5' 'Attribute6' 'Attribute7' 'Attribute0' 'Attribute1' 'Attribute2' 'Attribute3' 'Attribute4' 'Attribute5' 'Attribute6' 'Attribute7'

有人可以帮我吗？

Answer 1

将属性视为表的列，并且必须创建一次。

这是单维数据的代码。 在此示例中，我的假想表有一列为“attr1”，并且表有 其中有3条记录(实例)，结构非常简单易懂。

        Attribute attr1 = new Attribute("attr1");                               
        ArrayList<Attribute> attrList = new ArrayList<Attribute>();             
        attrList.add(attr1);                

        Instances dataset = new Instances("test", attrList, 0);

        double[] val1 = new double[] { 1.2};
        double[] val2 = new double[] { 2.2};
        double[] val3 = new double[] { 1.4};

        Instance instance0 = new DenseInstance(1.0, val1);
        instance0.setDataset(dataset);

        Instance instance1 = new DenseInstance(1.0, val2);
        instance1.setDataset(dataset);

        Instance instance2 = new DenseInstance(1.0, val3);
        instance2.setDataset(dataset);

        dataset.add(instance0);     
        dataset.add(instance1); 
        dataset.add(instance2); 

        SimpleKMeans kmeans = new SimpleKMeans();               
        try {
            kmeans.setPreserveInstancesOrder(true);
            kmeans.setNumClusters(2);
            kmeans.setSeed(2);
            kmeans.setDontReplaceMissingValues(true);
            kmeans.buildClusterer(dataset);
            kmeans.setMaxIterations(10);                                    
            Instances instances = kmeans.getClusterCentroids();
            int assignments[] = kmeans.getAssignments();
            int x=0;
            for(int assignment : assignments) {
                System.out.println("data :" + dataset.get(x) + "instance idx: " + x + " centroid value: " + instances.get(assignment));
                x++;
            }
         }