Skip to content

Add multiclassification topk and confmatrix metrics to model insights serialization format #537

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Jan 15, 2021
Merged

Add multiclassification topk and confmatrix metrics to model insights serialization format #537

Show file tree
Hide file tree
Changes from 4 commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
7c4b3e0
add topk and confusion matrix metrics to model insights serialization…
feifjiang Jan 5, 2021
abc240d
modified threshold for flaky RecordInisghtsLOCOTest
feifjiang Jan 5, 2021
e037c3f
(1)add case class ClassCount to represent MisClassifications (2)add t…
feifjiang Jan 8, 2021
33184bf
Merge branch 'master' into ff/insights
tovbinm Jan 8, 2021
9ea7e44
remove unused library imports
feifjiang Jan 8, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 3 additions & 1 deletion core/src/main/scala/com/salesforce/op/ModelInsights.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -401,7 +401,9 @@ case object ModelInsights {
classOf[DataBalancerSummary], classOf[DataCutterSummary], classOf[DataSplitterSummary],
classOf[SingleMetric], classOf[MultiMetrics], classOf[BinaryClassificationMetrics],
classOf[BinaryClassificationBinMetrics], classOf[MulticlassThresholdMetrics],
classOf[BinaryThresholdMetrics], classOf[MultiClassificationMetrics], classOf[RegressionMetrics]
classOf[BinaryThresholdMetrics], classOf[MultiClassificationMetrics], classOf[RegressionMetrics],
classOf[MultiClassificationMetricsTopK],
classOf[MulticlassConfMatrixMetricsByThreshold], classOf[MisClassificationMetrics]
))
val evalMetricsSerializer = new CustomSerializer[EvalMetric](_ =>
( { case JString(s) => EvalMetric.withNameInsensitive(s) },
Expand Down
92 changes: 51 additions & 41 deletions core/src/main/scala/com/salesforce/op/evaluators/OpMultiClassificationEvaluator.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -278,6 +278,34 @@ private[op] class OpMultiClassificationEvaluator
)
}

/**
* function to convert a sequence of ClassCount to a MisClassificationsPerCategory instance
*
* @param allClassCtSeq sequence of ClassCount containing labels or predictions and their counts for each category
* @param category index of a labelled or predicted class
* @return a MisClassifcationPerCategory instance
*/
private def getMisclassificationsPerCategory(
category: Double, allClassCtSeq: Seq[ClassCount]): MisClassificationsPerCategory = {

val misClassificationCtMap = allClassCtSeq
.filter(_.ClassIndex != category)
.sortBy(-_.Count)
.take($(confMatrixMinSupport))

val labelCount = allClassCtSeq.map(_.Count).reduce(_ + _)
val correctCount = allClassCtSeq.filter(_.ClassIndex == category)
.map(_.Count)
.reduceOption(_ + _).getOrElse(0L)

MisClassificationsPerCategory(
Category = category,
TotalCount = labelCount,
CorrectCount = correctCount,
MisClassifications = misClassificationCtMap
)
}

/**
* function to calculate the mostly frequently mis-classified classes for each label/prediction category
*
Expand All @@ -291,49 +319,15 @@ private[op] class OpMultiClassificationEvaluator
.reduceByKey(_ + _)

val misClassificationsByLabel = labelPredictionCountRDD.map {
case ((label, prediction), count) => (label, Seq((prediction, count)))
case ((label, prediction), count) => (label, Seq(ClassCount(prediction, count)))
}.reduceByKey(_ ++ _)
.map { case (label, predictionCountsIter) => {
val misClassificationCtMap = predictionCountsIter
.filter { case (pred, _) => pred != label }
.sortBy(-_._2)
.take($(confMatrixMinSupport)).toMap

val labelCount = predictionCountsIter.map(_._2).reduce(_ + _)
val correctCount = predictionCountsIter
.collect { case (pred, count) if pred == label => count }
.reduceOption(_ + _).getOrElse(0L)

MisClassificationsPerCategory(
Category = label,
TotalCount = labelCount,
CorrectCount = correctCount,
MisClassifications = misClassificationCtMap
)
}
}.sortBy(-_.TotalCount).collect()
.map { case (label, predictionCountsSeq) => getMisclassificationsPerCategory(label, predictionCountsSeq)}
.sortBy(-_.TotalCount).collect()

val misClassificationsByPrediction = labelPredictionCountRDD.map {
case ((label, prediction), count) => (prediction, Seq((label, count)))
case ((label, prediction), count) => (prediction, Seq(ClassCount(label, count)))
}.reduceByKey(_ ++ _)
.map { case (prediction, labelCountsIter) => {
val sortedMisclassificationCt = labelCountsIter
.filter { case (label, _) => label != prediction }
.sortBy(-_._2)
.take($(confMatrixMinSupport)).toMap

val predictionCount = labelCountsIter.map(_._2).reduce(_ + _)
val correctCount = labelCountsIter
.collect { case (label, count) if label == prediction => count }
.reduceOption(_ + _).getOrElse(0L)

MisClassificationsPerCategory(
Category = prediction,
TotalCount = predictionCount,
CorrectCount = correctCount,
MisClassifications = sortedMisclassificationCt
)
}
.map { case (prediction, labelCountsSeq) => getMisclassificationsPerCategory(prediction, labelCountsSeq)
}.sortBy(-_.TotalCount).collect()

MisClassificationMetrics(
Expand Down Expand Up @@ -541,10 +535,15 @@ case class MultiClassificationMetrics
*/
case class MultiClassificationMetricsTopK
(
@JsonDeserialize(contentAs = classOf[java.lang.Integer])
topKs: Seq[Int],
@JsonDeserialize(contentAs = classOf[java.lang.Double])
Precision: Seq[Double],
@JsonDeserialize(contentAs = classOf[java.lang.Double])
Recall: Seq[Double],
@JsonDeserialize(contentAs = classOf[java.lang.Double])
F1: Seq[Double],
@JsonDeserialize(contentAs = classOf[java.lang.Double])
Error: Seq[Double]
) extends EvaluationMetrics

Expand Down Expand Up @@ -594,8 +593,19 @@ case class MisClassificationsPerCategory
Category: Double,
TotalCount: Long,
CorrectCount: Long,
@JsonDeserialize(keyAs = classOf[java.lang.Double])
MisClassifications: Map[Double, Long]
MisClassifications: Seq[ClassCount]
)

/**
* container to store the count of a class
*
* @param ClassIndex
* @param Count
*/
case class ClassCount
(
ClassIndex: Double,
Count: Long
)

/**
Expand Down
59 changes: 57 additions & 2 deletions core/src/test/scala/com/salesforce/op/ModelInsightsTest.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -40,8 +40,8 @@ import com.salesforce.op.stages.impl.preparators._
import com.salesforce.op.stages.impl.regression.{OpLinearRegression, OpXGBoostRegressor, RegressionModelSelector}
import com.salesforce.op.stages.impl.selector.ModelSelectorNames.EstimatorType
import com.salesforce.op.stages.impl.selector.ValidationType._
import com.salesforce.op.stages.impl.selector.{SelectedCombinerModel, SelectedModel, SelectedModelCombiner}
import com.salesforce.op.stages.impl.tuning.{DataCutter, DataSplitter}
import com.salesforce.op.stages.impl.selector.{ModelEvaluation, ModelSelectorSummary, ProblemType, SelectedCombinerModel, SelectedModel, SelectedModelCombiner, ValidationType}
import com.salesforce.op.stages.impl.tuning.{DataBalancerSummary, DataCutter, DataSplitter}
import com.salesforce.op.test.{PassengerSparkFixtureTest, TestFeatureBuilder}
import com.salesforce.op.testkit.RandomReal
import com.salesforce.op.utils.spark.{OpVectorColumnMetadata, OpVectorMetadata}
Expand Down Expand Up @@ -406,6 +406,61 @@ class ModelInsightsTest extends FlatSpec with PassengerSparkFixtureTest with Dou
pretty should include("Top Contributions")
}

it should "correctly serialize and deserialize from json with MulticlassificationMetrics" in {
val trainMetrics = MultiClassificationMetrics(
Precision = 0.1,
Recall = 0.2,
F1 = 0.3,
Error = 0.4,
ThresholdMetrics = MulticlassThresholdMetrics(topNs = Seq(1, 2), thresholds = Seq(1.1, 1.2),
correctCounts = Map(1 -> Seq(100L)), incorrectCounts = Map(2 -> Seq(200L)),
noPredictionCounts = Map(3 -> Seq(300L))),
TopKMetrics = MultiClassificationMetricsTopK(Seq(1), Seq(0.1), Seq(0.1), Seq(0.1), Seq(0.1)),
ConfusionMatrixMetrics = MulticlassConfMatrixMetricsByThreshold( 2, Seq(0.1), Seq(0.1), Seq(Seq(1L))),
MisClassificationMetrics = MisClassificationMetrics(1, Seq.empty,
Seq(MisClassificationsPerCategory(0.0, 5L, 5L, Seq(ClassCount(1.0, 3L)))))
)

val holdoutMetrics = MultiClassificationMetrics(
Precision = 0.1,
Recall = 0.2,
F1 = 0.3,
Error = 0.4,
ThresholdMetrics = MulticlassThresholdMetrics(topNs = Seq(1, 2), thresholds = Seq(1.1, 1.2),
correctCounts = Map(1 -> Seq(100L)), incorrectCounts = Map(2 -> Seq(200L)),
noPredictionCounts = Map(3 -> Seq(300L))),
TopKMetrics = MultiClassificationMetricsTopK(Seq.empty, Seq.empty, Seq.empty, Seq.empty, Seq.empty),
ConfusionMatrixMetrics = MulticlassConfMatrixMetricsByThreshold(2, Seq(0.1), Seq(0.1), Seq.empty),
MisClassificationMetrics = MisClassificationMetrics(1, Seq.empty,
Seq(MisClassificationsPerCategory(0.0, 5L, 5L, Seq.empty)))
)

val summary = ModelSelectorSummary(
validationType = ValidationType.TrainValidationSplit,
validationParameters = Map.empty,
dataPrepParameters = Map.empty,
dataPrepResults = None,
evaluationMetric = MultiClassEvalMetrics.Error,
problemType = ProblemType.MultiClassification,
bestModelUID = "test1",
bestModelName = "test2",
bestModelType = "test3",
validationResults = Seq.empty,
trainEvaluation = trainMetrics,
holdoutEvaluation = Some(holdoutMetrics)
)

val insights = workflowModel.modelInsights(pred).copy(selectedModelInfo = Some(summary))
ModelInsights.fromJson(insights.toJson()) match {
case Failure(e) => fail(e)
case Success(deser) =>
insights.selectedModelInfo.toSeq.zip(deser.selectedModelInfo.toSeq).foreach {
case (o, i) =>
o.trainEvaluation shouldEqual i.trainEvaluation
o.holdoutEvaluation shouldEqual i.holdoutEvaluation
}
}
}

it should "correctly serialize and deserialize from json when raw feature filter is not used" in {
val insights = workflowModel.modelInsights(pred)
Expand Down
28 changes: 14 additions & 14 deletions core/src/test/scala/com/salesforce/op/evaluators/OpMultiClassificationEvaluatorTest.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -398,10 +398,10 @@ class OpMultiClassificationEvaluatorTest extends FlatSpec with TestSparkContext

// create a test 2D array where 1st dimension is the label and 2nd dimension is the prediction,
// and the # of (label, prediction) equals to the value of the label
// _| 1 2 3
// 1| 1L 1L 1L
// 2| 2L 2L 2L
// 3| 3L 3L 3L
// ___| 1.0 2.0 3.0
// 1.0| 1L 1L 1L
// 2.0| 2L 2L 2L
// 3.0| 3L 3L 3L
val testLabels = Array(1.0, 2.0, 3.0)
val labelAndPrediction = testLabels.flatMap(label => {
testLabels.flatMap(pred => Seq.fill(label.toInt)((label, pred)))
Expand Down Expand Up @@ -437,10 +437,10 @@ class OpMultiClassificationEvaluatorTest extends FlatSpec with TestSparkContext

// create a test 2D array with the count of each label & prediction combination as:
// row is label and column is prediction
// _| 1 2 3
// 1| 2L 3L 4L
// 2| 3L 4L 5L
// 3| 4L 5L 6L
// ___| 1.0 2.0 3.0
// 1.0| 2L 3L 4L
// 2.0| 3L 4L 5L
// 3.0| 4L 5L 6L
val testLabels = List(1.0, 2.0, 3.0)
val labelAndPrediction = testLabels.flatMap(label => {
testLabels.flatMap(pred => Seq.fill(label.toInt + pred.toInt)((label, pred)))
Expand All @@ -452,21 +452,21 @@ class OpMultiClassificationEvaluatorTest extends FlatSpec with TestSparkContext
outputMetrics.MisClassificationsByLabel shouldEqual
Seq(
MisClassificationsPerCategory(Category = 3.0, TotalCount = 15L, CorrectCount = 6L,
MisClassifications = Map(2.0 -> 5L, 1.0 -> 4L)),
MisClassifications = Seq(ClassCount(2.0, 5L), ClassCount(1.0, 4L))),
MisClassificationsPerCategory(Category = 2.0, TotalCount = 12L, CorrectCount = 4L,
MisClassifications = Map(3.0 -> 5L, 1.0 -> 3L)),
MisClassifications = Seq(ClassCount(3.0, 5L), ClassCount(1.0, 3L))),
MisClassificationsPerCategory(Category = 1.0, TotalCount = 9L, CorrectCount = 2L,
MisClassifications = Map(3.0 -> 4L, 2.0 -> 3L))
MisClassifications = Seq(ClassCount(3.0, 4L), ClassCount(2.0, 3L)))
)

outputMetrics.MisClassificationsByPrediction shouldEqual
Seq(
MisClassificationsPerCategory(Category = 3.0, TotalCount = 15L, CorrectCount = 6L,
MisClassifications = Map(2.0 -> 5L, 1.0 -> 4L)),
MisClassifications = Seq(ClassCount(2.0, 5L), ClassCount(1.0, 4L))),
MisClassificationsPerCategory(Category = 2.0, TotalCount = 12L, CorrectCount = 4L,
MisClassifications = Map(3.0 -> 5L, 1.0 -> 3L)),
MisClassifications = Seq(ClassCount(3.0, 5L), ClassCount(1.0, 3L))),
MisClassificationsPerCategory(Category = 1.0, TotalCount = 9L, CorrectCount = 2L,
MisClassifications = Map(3.0 -> 4L, 2.0 -> 3L))
MisClassifications = Seq(ClassCount(3.0, 4L), ClassCount(2.0, 3L)))
)
}
}
Expand Down
4 changes: 2 additions & 2 deletions core/src/test/scala/com/salesforce/op/stages/impl/insights/RecordInsightsLOCOTest.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -232,8 +232,8 @@ class RecordInsightsLOCOTest extends FunSpec with TestSparkContext with RecordIn
info("Each feature vector should only have either three or four non-zero entries. One each from country and " +
"picklist, while currency can have either two (if it's null the currency column will be filled with the mean)" +
" or just one if it's not null.")
it("should pick between 1 and 4 of the features") {
all(parsed.map(_.size)) should (be >= 1 and be <= 4)
it("should pick between 0 and 4 of the features") {
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why did this change?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@tovbinm this test is quite flaky. In one of the failed runs, the value was 0 and not between 1 and 4.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@leahmcguire @Jauntbox do you have any ideas?

Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

There's still an open PR to address this: #423

We could use a better test, but for now we can start by getting rid of the flakiness.

all(parsed.map(_.size)) should (be >= 0 and be <= 4)
}

// Grab the feature vector metadata for comparison against the LOCO record insights
Expand Down
18 changes: 11 additions & 7 deletions core/src/test/scala/com/salesforce/op/stages/impl/selector/ModelSelectorSummaryTest.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -97,10 +97,10 @@ class ModelSelectorSummaryTest extends FlatSpec with TestSparkContext {
correctCounts = Map(1 -> Seq(100L)), incorrectCounts = Map(2 -> Seq(200L)),
noPredictionCounts = Map(3 -> Seq(300L))),
TopKMetrics = MultiClassificationMetricsTopK(Seq.empty, Seq.empty, Seq.empty, Seq.empty, Seq.empty),
ConfusionMatrixMetrics = MulticlassConfMatrixMetricsByThreshold(1, Seq(1.0), Seq(0.0, 0.5), Seq.empty),
ConfusionMatrixMetrics = MulticlassConfMatrixMetricsByThreshold(1, Seq(1.0), Seq(0.0, 0.5), Seq(Seq(1L))),
MisClassificationMetrics = MisClassificationMetrics(1, Seq.empty,
Seq(MisClassificationsPerCategory(TotalCount = 5L, CorrectCount = 3L, Category = 1.0,
MisClassifications = Map(1.0 -> 2L))))),
MisClassifications = Seq(ClassCount(1.0, 2L)))))),
holdoutEvaluation = None
)

Expand All @@ -121,19 +121,23 @@ class ModelSelectorSummaryTest extends FlatSpec with TestSparkContext {
}

it should "not hide the root cause of JSON parsing errors" in {
val evalMetrics = MultiClassificationMetrics(Precision = 0.1, Recall = 0.2, F1 = 0.3, Error = 0.4,
val evalMetrics = MultiClassificationMetrics(
Precision = 0.1,
Recall = 0.2,
F1 = 0.3,
Error = 0.4,
ThresholdMetrics = MulticlassThresholdMetrics(topNs = Seq(1, 2), thresholds = Seq(1.1, 1.2),
correctCounts = Map(1 -> Seq(100L)), incorrectCounts = Map(2 -> Seq(200L)),
noPredictionCounts = Map(3 -> Seq(300L))),
TopKMetrics = MultiClassificationMetricsTopK(Seq.empty, Seq.empty, Seq.empty, Seq.empty, Seq.empty),
ConfusionMatrixMetrics = MulticlassConfMatrixMetricsByThreshold( 2, Seq(0.1), Seq(0.1),
Seq.empty),
ConfusionMatrixMetrics = MulticlassConfMatrixMetricsByThreshold( 2, Seq(0.1), Seq(0.1), Seq(Seq(1L))),
MisClassificationMetrics = MisClassificationMetrics(1,
Seq(MisClassificationsPerCategory(0.0, 5L, 5L, Map(1.0 -> 3L))),
Seq(MisClassificationsPerCategory(0.0, 5L, 5L, Map(1.0 -> 3L))))
Seq(MisClassificationsPerCategory(0.0, 5L, 5L, Seq(ClassCount(1.0, 3L)))),
Seq(MisClassificationsPerCategory(0.0, 5L, 5L, Seq(ClassCount(1.0, 3L)))))
)

val evalMetricsJson = evalMetrics.toJson()
println(1)
val roundTripEvalMetrics = ModelSelectorSummary.evalMetFromJson(
classOf[MultiClassificationMetrics].getName, evalMetricsJson).get
roundTripEvalMetrics shouldBe evalMetrics
Expand Down