CODEWITHSUNDEEP
javawekadecision-tree
Nipun Alahakoon
Nipun Alahakoon

Reputation: 2862

Weka J-48 decision Tree not completing

I am using a multi-attribute dataset for classification purpose. I am using WEKA API on java.The dataset have both categorical and numerical variables. When i run the dataset on weka-GUI i get a better result with 16 leaves in 26 sized tree. But when i do the same using java code i only get 3 leaves in 5 sized tree . Here is my java code

public static Evaluation classify(Classifier model,
        Instances trainingSet, Instances testingSet) throws Exception {

    //return the classification model after training with train set and test with test set

    Evaluation evaluation = new Evaluation(trainingSet);

    model.buildClassifier(trainingSet);
    evaluation.evaluateModel(model, testingSet);
    //System.out.println(model);
    return evaluation;
}

    Classifier models = new J48(); // a decision tree
    models.setOptions(optionsj);
    FastVector predictions = new FastVector();
    // For each training-testing split pair, train and test the classifier
    for (int i = 0; i < trainingSplits.length; i++) {
        Evaluation validation = classify(models, trainingSplits[i], testingSplits[i]);

        predictions.appendElements(validation.predictions());
        System.out.println(validation.toSummaryString("\nResults\n======\n", false));
    }

    System.out.println(models.toString());

How to make sure the j-48 take all the attributes in the dataset? what i did wrong?

Upvotes: 0

Views: 566

Answers (1)

Daniel Alexander
Daniel Alexander

Reputation: 26

there is a parameter you have to set which is called either pruned or non-pruned. pruned mean the tree only display the most important leaves of the decision tree. non-pruned mean it displayed every variable in the decision tree. here you are just using pruned tree. if you want the full tree set 

pruned = "False"

Upvotes: 1

Related Questions