zhengruifeng commented on a change in pull request #26135:
[SPARK-29489][ML][PySpark] ml.evaluation support log-loss
URL: https://github.com/apache/spark/pull/26135#discussion_r335327837
##########
File path:
mllib/src/main/scala/org/apache/spark/mllib/evaluation/MulticlassMetrics.scala
##########
@@ -28,28 +28,30 @@ import org.apache.spark.sql.{DataFrame, Row}
/**
* Evaluator for multiclass classification.
*
- * @param predictionAndLabels an RDD of (prediction, label, weight) or
- * (prediction, label) tuples.
+ * @param predictionAndLabels an RDD of (prediction, label, weight,
probability) or
+ * (prediction, label, weight) or (prediction,
label) tuples.
*/
@Since("1.1.0")
class MulticlassMetrics @Since("1.1.0") (predictionAndLabels: RDD[_ <:
Product]) {
/**
* An auxiliary constructor taking a DataFrame.
- * @param predictionAndLabels a DataFrame with two double columns:
prediction and label
+ * @param predictionAndLabels a DataFrame with columns: prediction, label,
weight(optional)
+ * and probability(only for logloss)
*/
private[mllib] def this(predictionAndLabels: DataFrame) =
- this(predictionAndLabels.rdd.map {
- case Row(prediction: Double, label: Double, weight: Double) =>
- (prediction, label, weight)
- case Row(prediction: Double, label: Double) =>
- (prediction, label, 1.0)
- case other =>
- throw new IllegalArgumentException(s"Expected Row of tuples, got
$other")
+ this(predictionAndLabels.rdd.map { r =>
Review comment:
matching will not work in pyspark, so I have to use `r.get` instead.
`MultilabelMetrics` also deals with dataframe in this way.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
