Package

com.salesforce.op.stages.impl

tuning

Permalink

package tuning

Visibility
  1. Public
  2. All

Type Members

  1. case class BestEstimator[E <: Estimator[_]](name: String, estimator: E, summary: Seq[ModelEvaluation]) extends Product with Serializable

    Permalink

    Best Estimator container

    Best Estimator container

    E

    model type

    name

    the name of the best model

    estimator

    best estimator

    summary

    optional metadata

  2. class DataBalancer extends Splitter with DataBalancerParams

    Permalink

    Instance that will split the data into train and holdout and then balance the dataset before modeling binary classifications

  3. trait DataBalancerParams extends Params

    Permalink
  4. case class DataBalancerSummary(positiveLabels: Long, negativeLabels: Long, desiredFraction: Double, upSamplingFraction: Double, downSamplingFraction: Double) extends SplitterSummary with Product with Serializable

    Permalink

    Summary for data balancer run for storage in metadata

    Summary for data balancer run for storage in metadata

    positiveLabels

    count of positive labels

    negativeLabels

    count of negative labels

    desiredFraction

    desired min fraction of smaller label count

    upSamplingFraction

    up/down sampling for smaller class of label

    downSamplingFraction

    down sampling for larger class of label

  5. class DataCutter extends Splitter with DataCutterParams

    Permalink

    Instance that will make a holdout set and prepare the data for multiclass modeling Creates instance that will split data into training and test set filtering out any labels that don't meet the minimum fraction cutoff or fall in the top N labels specified.

  6. case class DataCutterSummary(labelsKept: Seq[Double], labelsDropped: Seq[Double]) extends SplitterSummary with Product with Serializable

    Permalink

    Summary of results for data cutter

    Summary of results for data cutter

    labelsKept

    labels retained

    labelsDropped

    labels dropped by data cutter

  7. class DataSplitter extends Splitter

    Permalink

    Instance that will split the data into training and holdout for regressions

  8. case class DataSplitterSummary() extends SplitterSummary with Product with Serializable

    Permalink

    Empty class because no summary information for a data splitter

  9. case class ModelData(train: Dataset[Row], summary: Option[SplitterSummary]) extends Product with Serializable

    Permalink

    Case class for Training & test sets

    Case class for Training & test sets

    train

    training set is persisted at construction

    summary

    summary for building metadata

  10. abstract class Splitter extends SplitterParams

    Permalink

    Abstract class that will carry on the creation of training set + test set

  11. trait SplitterParams extends Params

    Permalink
  12. trait SplitterSummary extends MetadataLike

    Permalink

Value Members

  1. object DataBalancer extends Product with Serializable

    Permalink
  2. object DataCutter extends Product with Serializable

    Permalink
  3. object DataSplitter extends Product with Serializable

    Permalink
  4. object SplitterParamsDefault

    Permalink
  5. object ValidatorParamDefaults

    Permalink

Ungrouped