Example usage for weka.core Instances checkForAttributeType

Introduction

In this page you can find the example usage for weka.core Instances checkForAttributeType.

Prototype

public boolean checkForAttributeType(int attType) 

Source Link

Document

Checks for attributes of the given type in the dataset

Usage

From source file:org.opentox.qsar.processors.trainers.classification.WekaClassifier.java

License:Open Source License

@Override
public Instances preprocessData(Instances data) throws QSARException {
    /*//from   www. j a v  a2s . co m
     * TODO: In case a client choses a non-nominal feature for the classifier,
     * provide a list of some available nominal features.
     */

    if (data == null) {
        throw new NullPointerException("Cannot train a classification model without data");
    }

    /* The incoming dataset always has the first attribute set to
    'compound_uri' which is of type "String". This is removed at the
    begining of the training procedure */
    AttributeCleanup filter = new AttributeCleanup(ATTRIBUTE_TYPE.string);
    // NOTE: Removal of string attributes should be always performed prior to any kind of training!
    data = filter.filter(data);

    SimpleMVHFilter fil = new SimpleMVHFilter();
    data = fil.filter(data);

    // CHECK IF THE GIVEN URI IS AN ATTRIBUTE OF THE DATASET
    Attribute classAttribute = data.attribute(predictionFeature);
    if (classAttribute == null) {
        throw new QSARException(Cause.XQReg202,
                "The prediction feature you provided is not a valid numeric attribute of the dataset :{"
                        + predictionFeature + "}");
    }

    // CHECK IF THE DATASET CONTAINS ANY NOMINAL ATTRIBUTES
    if (!data.checkForAttributeType(Attribute.NOMINAL)) {
        throw new QSARException(Cause.XQC4040, "Improper dataset! The dataset you provided has no "
                + "nominal features therefore classification models cannot be built.");
    }

    // CHECK WHETHER THE CLASS ATTRIBUTE IS NOMINAL
    if (!classAttribute.isNominal()) {
        StringBuilder list_of_nominal_features = new StringBuilder();

        int j = 0;
        for (int i = 0; i < data.numAttributes() && j < 10; i++) {
            if (data.attribute(i).isNominal()) {
                j++;
                list_of_nominal_features.append(data.attribute(i).name() + "\n");
            }
        }

        throw new QSARException(Cause.XQC4041,
                "The prediction feature you provided "
                        + "is not a nominal. Here is a list of some nominal features in the dataset you might "
                        + "be interested in :\n" + list_of_nominal_features.toString());
    }

    // CHECK IF THE RANGE OF THE CLASS ATTRIBUTE IS NON-UNARY
    Enumeration nominalValues = classAttribute.enumerateValues();
    String v = nominalValues.nextElement().toString();
    if (!nominalValues.hasMoreElements()) {
        throw new QSARException(Cause.XQC4042,
                "This classifier cannot handle unary nominal classes, that is "
                        + "nominal class attributes whose range includes only one value. Singleton value : {"
                        + v + "}");
    }

    // SET THE CLASS ATTRIBUTE OF THE DATASET
    data.setClass(classAttribute);

    return data;
}

---