Github user jkbradley commented on a diff in the pull request:
https://github.com/apache/spark/pull/7538#discussion_r36361293
--- Diff:
mllib/src/main/scala/org/apache/spark/ml/classification/LogisticRegression.scala
---
@@ -286,6 +294,41 @@ class LogisticRegressionModel private[ml] (
override val numClasses: Int = 2
+ private var trainingSummary:
Option[BinaryLogisticRegressionTrainingSummary] = None
+
+ /**
+ * Gets summary of model on training set. An exception is
+ * thrown if `trainingSummary == None`.
+ */
+ def summary: BinaryLogisticRegressionTrainingSummary = trainingSummary
match {
+ case Some(summ) => summ
+ case None =>
+ throw new SparkException(
+ "No training summary available for this LogisticRegressionModel",
+ new NullPointerException())
+ }
+
+ private[classification] def setSummary(
+ summary: BinaryLogisticRegressionTrainingSummary): this.type = {
+ this.trainingSummary = Some(summary)
+ this
+ }
+
+ /** Indicates whether a training summary exists for this model instance.
*/
+ def hasSummary: Boolean = trainingSummary.isDefined
+
+ /**
+ * Evaluates the model on a testset.
+ * @param dataset Test dataset to evaluate model on.
+ */
+ // TODO: decide on a good name before exposing to public API
+ private[classification] def evaluate(dataset: DataFrame):
BinaryLogisticRegressionSummary = {
+ val t = udf { features: Vector =>
raw2probabilityInPlace(predictRaw(features)) }
+ val labelsAndScores = dataset.
--- End diff --
It'd be better just to call ```this.transform(dataset)``` and pass that to
the summary constructor. Less code, and it keeps more columns.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
