exploration

Type Members

1. class FeatureImportances extends FeatureImportanceTools with SparkSessionWrapper

   
   Example:

   1. val genericMapOverrides = Map("labelCol" -> "label", "tunerKFold" -> 2, "tunerTrainSplitMethod" -> "stratified", "tunerNumberOfGenerations" -> 4, "tunerNumberOfMutationsPerGeneration" -> 6, "tunerInitialGenerationPermutationCount" -> 25, "fieldsToIgnoreInVector" -> Array("final_weight"),"tunerInitialGenerationMode" -> "permutations") val xgbConfig = ConfigurationGenerator.generateConfigFromMap("XGBoost", "classifier", genericMapOverrides) val featConfig = ConfigurationGenerator.generateFeatureImportanceConfig(xgbConfig) val importances = new FeatureImportances(data, featConfig, "count", 5.0).generateFeatureImportances()

   Note

   for cutoffType 'None', this value can be set to 0.0

   ,

   Count => Return the top n most important fields maxing the naming of the original data, in descending order, are returned.
   None => A sorted list of columns, in descending order of importance, are returned.
   Value => All values above the thresholded value set in cutoffValue are returned in descending order.