Java Code Examples for org.apache.spark.mllib.evaluation.MulticlassMetrics#labels()

@Override
public double evaluate(DataFrame dataset) {
    StructType schema = dataset.schema();

    SchemaUtils.checkColumnType(schema, this.getPredictionCol(), DataTypes.DoubleType, "");
    SchemaUtils.checkColumnType(schema, this.getLabelCol(), DataTypes.DoubleType, "");

    MulticlassMetrics metrics = new MulticlassMetrics(dataset
            .select(this.getPredictionCol(), this.getLabelCol()));


    int labelColumn = 0;
    for(int i=0; i < metrics.labels().length; i++) {
        if(metrics.labels()[i] == evaluationClass) {
            labelColumn = i;
        }
    }

    double metric=0d;
    switch(getMetricName()) {
        case "f1":
            metric = metrics.fMeasure(metrics.labels()[labelColumn]);
            break;
        case "precision":
            metric = metrics.precision(metrics.labels()[labelColumn]);
            break;
        case "recall":
            metric = metrics.recall(metrics.labels()[labelColumn]);
            break;
        case "weightedPrecision":
            metric = metrics.weightedPrecision();
            break;
        case "weightedRecall":
            metric = metrics.weightedRecall();
            break;


    }

    return metric;
}
 

private void multiClassEvaluation(DataFrame predictions, String output, TrainingSettings trainingSettings) throws IOException {
    FileSystem fs = FileSystem.get(new Configuration());
    Path evalPath = new Path(output+"multiclass_evaluation_"+trainingSettings.getClassificationMethod()+".txt");
    fs.delete(evalPath, true);
    FSDataOutputStream fsdos = fs.create(evalPath);

    MulticlassMetrics metrics = new MulticlassMetrics(predictions
            .select("prediction", "label"));

    // Confusion matrix
    Matrix confusion = metrics.confusionMatrix();
    IOUtils.write("\nConfusion matrix: \n" + confusion, fsdos);

    // Overall statistics
    IOUtils.write("\nPrecision = " + metrics.precision(), fsdos);
    IOUtils.write("\nRecall = " + metrics.recall(), fsdos);
    IOUtils.write("\nF1 Score = " + metrics.fMeasure(), fsdos);
    IOUtils.write("\n\n", fsdos);
    // Stats by labels
    for (int i = 0; i < metrics.labels().length; i++) {


        IOUtils.write(String.format("Class %f precision = %f\n", metrics.labels()[i],metrics.precision(metrics.labels()[i])), fsdos);
        IOUtils.write(String.format("Class %f recall = %f\n", metrics.labels()[i], metrics.recall(metrics.labels()[i])), fsdos);
        IOUtils.write(String.format("Class %f F1 score = %f\n", metrics.labels()[i], metrics.fMeasure(metrics.labels()[i])), fsdos);

        System.out.format("Class %f precision = %f\n", metrics.labels()[i],metrics.precision(metrics.labels()[i]));
        System.out.format("Class %f recall = %f\n", metrics.labels()[i], metrics.recall(metrics.labels()[i]));
        System.out.format("Class %f F1 score = %f\n", metrics.labels()[i], metrics.fMeasure(metrics.labels()[i]));
    }

    //Weighted stats
    IOUtils.write("\nWeighted precision = "+metrics.weightedPrecision(), fsdos);
    IOUtils.write("\nWeighted recall = "+metrics.weightedRecall(), fsdos);
    IOUtils.write("\nWeighted F1 score ="+metrics.weightedFMeasure(), fsdos);
    IOUtils.write("\nWeighted false positive rate = " +metrics.weightedFalsePositiveRate(), fsdos);

    fsdos.flush();
    IOUtils.closeQuietly(fsdos);

}
 

public static void main(String[] args) {
  SparkConf conf = new SparkConf().setAppName("Multi class Classification Metrics Example");
  SparkContext sc = new SparkContext(conf);
  // $example on$
  String path = "data/mllib/sample_multiclass_classification_data.txt";
  JavaRDD<LabeledPoint> data = MLUtils.loadLibSVMFile(sc, path).toJavaRDD();

  // Split initial RDD into two... [60% training data, 40% testing data].
  JavaRDD<LabeledPoint>[] splits = data.randomSplit(new double[]{0.6, 0.4}, 11L);
  JavaRDD<LabeledPoint> training = splits[0].cache();
  JavaRDD<LabeledPoint> test = splits[1];

  // Run training algorithm to build the model.
  final LogisticRegressionModel model = new LogisticRegressionWithLBFGS()
    .setNumClasses(3)
    .run(training.rdd());

  // Compute raw scores on the test set.
  JavaRDD<Tuple2<Object, Object>> predictionAndLabels = test.map(
    new Function<LabeledPoint, Tuple2<Object, Object>>() {
      public Tuple2<Object, Object> call(LabeledPoint p) {
        Double prediction = model.predict(p.features());
        return new Tuple2<Object, Object>(prediction, p.label());
      }
    }
  );

  // Get evaluation metrics.
  MulticlassMetrics metrics = new MulticlassMetrics(predictionAndLabels.rdd());

  // Confusion matrix
  Matrix confusion = metrics.confusionMatrix();
  System.out.println("Confusion matrix: \n" + confusion);

  // Overall statistics
  System.out.println("Accuracy = " + metrics.accuracy());

  // Stats by labels
  for (int i = 0; i < metrics.labels().length; i++) {
    System.out.format("Class %f precision = %f\n", metrics.labels()[i],metrics.precision(
      metrics.labels()[i]));
    System.out.format("Class %f recall = %f\n", metrics.labels()[i], metrics.recall(
      metrics.labels()[i]));
    System.out.format("Class %f F1 score = %f\n", metrics.labels()[i], metrics.fMeasure(
      metrics.labels()[i]));
  }

  //Weighted stats
  System.out.format("Weighted precision = %f\n", metrics.weightedPrecision());
  System.out.format("Weighted recall = %f\n", metrics.weightedRecall());
  System.out.format("Weighted F1 score = %f\n", metrics.weightedFMeasure());
  System.out.format("Weighted false positive rate = %f\n", metrics.weightedFalsePositiveRate());

  // Save and load model
  model.save(sc, "target/tmp/LogisticRegressionModel");
  LogisticRegressionModel sameModel = LogisticRegressionModel.load(sc,
    "target/tmp/LogisticRegressionModel");
  // $example off$
}