weka特征选择之构建数据集

1、weka数据集基本概念 

weka.core.Instances-拥有一个完整的数据集。这个数据结构是基于行的；单一的行可以通过从0开始索引的instance(int) 方法获得。有关列的信息可以通过attribute(int)方法获得。此方法返回 weka.core.Attribute对象。

 weka.core.Instance。-封装了一个单行。它基本上是一个双原语数组的包装。因为这个类不包含关于列的类型的任何信息，它总是需要访问weka.core.Instances对象。

weka.core.Attribute -拥有数据集中单个列的类型信息。它存储属性的类型，以及标称属性的标签，字符串属性可能的值或关系属性的数据集（这些也都是weka.core.Instances对象）。

2、手动构建weka数据输入Instances

    public Instances produceData(){

       Attribute length = new Attribute("length");

       Attribute weight = new Attribute("weight");

      

       // Create vector to hold nominal values "first", "second", "third"

       FastVector my_nominal_values = new FastVector(3);

       my_nominal_values.addElement("first");

       my_nominal_values.addElement("second");

       my_nominal_values.addElement("third");

      

       // Create nominal attribute "position"

       Attribute position = new Attribute("position", my_nominal_values);

      

       // Create vector of the above attributes

       FastVector attributes = new FastVector(3);

       attributes.addElement(length);

       attributes.addElement(weight);

       attributes.addElement(position);

      

       // Create the empty dataset "race" with above attributes

       Instances race = new Instances("race", attributes, 0);

      

       // Make position the class attribute

       race.setClassIndex(position.index());

      

       // Create empty instance with three attribute values

       Instance inst = new Instance(3);

      

       // Set instance's values for the attributes "length", "weight", and "position"

       inst.setValue(length, 5.3);

       inst.setValue(weight, 300);

       inst.setValue(position, "first");

      

       // Set instance's dataset to be the dataset "race"

       inst.setDataset(race);

       race.add(inst);

       return race;

    }

3、对输入的数据集Instances进行特征选择

    private ArffLoader loader;

    private Instances dataSet;

    private File arffFile;

    private int sizeOfDataset;

    private int numOfOldAttributes;

    private int numOfNewAttributes;

    private int classIndex;

    private int[] selectedAttributes;

 

    public WekaSelector(File file) throws IOException {

        loader = new ArffLoader();

        arffFile = file;

        loader.setFile(arffFile);

        dataSet = loader.getDataSet();

        sizeOfDataset = dataSet.numInstances();

        numOfOldAttributes = dataSet.numAttributes();

        classIndex = numOfOldAttributes - 1;

        dataSet.setClassIndex(classIndex);

    }

 

    public void select() throws Exception {

        ASEvaluation evaluator = new CfsSubsetEval();

        ASSearch search = new BestFirst();

        AttributeSelection eval = null;

 

        eval = new AttributeSelection();

        eval.setEvaluator(evaluator);

        eval.setSearch(search);

        

        Instances data = produceData();

        //eval.SelectAttributes(dataSet);

        eval.SelectAttributes(data);

       

        numOfNewAttributes = eval.numberAttributesSelected();

        selectedAttributes = eval.selectedAttributes();

        System.out.println("result is "+eval.toResultsString());

       // System.out.println("old number of Attributes is "+numOfOldAttributes);

        System.out.println("new number of Attributes is "+numOfNewAttributes);

        for(int i=0;i<selectedAttributes.length;i++){

            System.out.println(selectedAttributes[i]);

        }

    }

 

 

转载于:https://my.oschina.net/u/3437790/blog/1525154