An evaluator that produces multiclass classification metrics given predicted and actual classes,
derived from computing the confusion matrix associated with these two sets of values.
Number of classes present.
Builds the confusion matrix for a multiclass classifier,
and provides common metrics such as micro & macro precision & recall
Similar to MLlib's org.apache.spark.mllib.evaluation.MulticlassMetrics,
but only does one pass over the data to calculate everything
An RDD of predicted class labels. Must range from 0 until numClasses
An RDD of the true class labels. Must range from 0 until numClasses
Common multiclass classifier metrics for this data