PostProcessor.h File Reference

#include <vector>
#include <SerializableObject/SerializableObject/SerializableObject.h>
#include "MedSamples.h"
#include "MedModel.h"
#include "Effected_Field.h"

Go to the source code of this file.

Data Structures
class	PostProcessor
	An Abstract PostProcessor class. More...
class	MultiPostProcessor
	A wrapper for parallel call to post_processors group. More...

Enumerations
enum	PostProcessorTypes {   FTR_POSTPROCESS_MULTI , FTR_POSTPROCESS_CALIBRATOR , FTR_POSTPROCESS_TREE_SHAP , FTR_POSTPROCESS_SHAPLEY ,   FTR_POSTPROCESS_MISSING_SHAP , FTR_POSTPROCESS_LIME_SHAP , FTR_POSTPROCESS_KNN_EXPLAIN , FTR_POSTPROCESS_LINEAR ,   FTR_POSTPROCESS_ITERATIVE_SET , FTR_POSTPROCESS_AGGREGATE_PREDS , FTR_POSTPROCESS_ADJUST , FTR_POSTPROCESS_FAIRNESS ,   FTR_POSTPROCESS_LAST }

Functions
PostProcessorTypes	post_processor_name_to_type (const string &post_processor)

Enumeration Type Documentation

PostProcessorTypes

enum PostProcessorTypes

Enumerator
FTR_POSTPROCESS_MULTI	"multi_processor" or "multi" to create MultiPostProcessor
FTR_POSTPROCESS_CALIBRATOR	"calibrator" to create Calibrator
FTR_POSTPROCESS_TREE_SHAP	"tree_shap" to create TreeExplainer to explain tree mode or mimic generic model with trees model
FTR_POSTPROCESS_SHAPLEY	"shapley" to create ShapleyExplainer - model agnostic shapley explainer for model. sample masks using gibbs or GAN
FTR_POSTPROCESS_MISSING_SHAP	"missing_shap" to create MissingShapExplainer - model agnostic shapley algorithm on trained model to handle prediciton with missing values(retrains new model). much faster impl, because gibbs/GAN is not needed
FTR_POSTPROCESS_LIME_SHAP	"lime_shap" to create LimeExplainer - model agnostic shapley algorithm with lime on shap values sampling
FTR_POSTPROCESS_KNN_EXPLAIN	"knn" Explainer built on knn principles KNN_Explainer
FTR_POSTPROCESS_LINEAR	"linear" to create LinearExplainer to explain linear model - importance is score change when putting zero in the feature/group of features
FTR_POSTPROCESS_ITERATIVE_SET	"iterative_set" to create IterativeSetExplainer - model agnostic iterative explainer for model. sample masks using gibbs or GAN
FTR_POSTPROCESS_AGGREGATE_PREDS	"aggregate_preds" to create AggregatePredsPostProcessor - averaging model predictions after resampling
FTR_POSTPROCESS_ADJUST	"adjust_probs" to adjust model calibrated predictions according to priors. Creates ProbAdjustPostProcessor
FTR_POSTPROCESS_FAIRNESS	"fairness_adjust" to adjust model calibrated predictions according to priors. Creates ProbAdjustPostProcessor