MethodC50.h

Go to the documentation of this file.

// @(#)root/tmva/rmva $Id$
// Author: Omar Zapata,Lorenzo Moneta, Sergei Gleyzer 2015
 
/**********************************************************************************
 * Project: TMVA - a Root-integrated toolkit for multivariate data analysis       *
 * Package: TMVA                                                                  *
 * Class  : RMethodC50                                                            *
 *                                                                                *
 * Description:                                                                   *
 *      R´s Package C50  method based on ROOTR                                    *
 *                                                                                *
 **********************************************************************************/
 
#ifndef ROOT_TMVA_RMethodC50
#define ROOT_TMVA_RMethodC50
 
//////////////////////////////////////////////////////////////////////////
//                                                                      //
// RMethodC50                                                          //
//                                                                      //
//                                                                      //
//////////////////////////////////////////////////////////////////////////
 
#include "TMVA/RMethodBase.h"
#include <vector>
 
namespace TMVA {
 
   class Factory;  // DSMTEST
   class Reader;   // DSMTEST
   class DataSetManager;  // DSMTEST
   class Types;
   class MethodC50 : public RMethodBase {
 
   public :
 
      // constructors
      MethodC50(const TString &jobName,
                const TString &methodTitle,
                DataSetInfo &theData,
                const TString &theOption = "");
 
      MethodC50(DataSetInfo &dsi,
                const TString &theWeightFile);
 
 
      ~MethodC50(void);
      void Train() override;
      // options treatment
      void Init() override;
      void DeclareOptions() override;
      void ProcessOptions() override;
      // create ranking
      const Ranking *CreateRanking() override
      {
         return nullptr;  // = 0;
      }
 
 
      Bool_t HasAnalysisType(Types::EAnalysisType type, UInt_t numberClasses, UInt_t numberTargets) override;
 
      // performs classifier testing
      void TestClassification() override;
 
 
      Double_t GetMvaValue(Double_t *errLower = nullptr, Double_t *errUpper = nullptr) override;
      void     MakeClass(const TString &classFileName = TString("")) const override;  //required for model persistence
      using MethodBase::ReadWeightsFromStream;
      // the actual "weights"
      void AddWeightsXMLTo(void * /*parent*/) const override {} // = 0;
      void ReadWeightsFromXML(void * /*weight*/) override {} // = 0;
      void ReadWeightsFromStream(std::istream &) override {} //= 0;       // backward compatibility
 
      // signal/background classification response for all current set of data
      std::vector<Double_t> GetMvaValues(Long64_t firstEvt = 0, Long64_t lastEvt = -1, Bool_t logProgress = false) override;
 
      void ReadModelFromFile();
   private :
      DataSetManager    *fDataSetManager;     // DSMTEST
      friend class Factory;                   // DSMTEST
      friend class Reader;                    // DSMTEST
   protected:
      //C5.0 function options
      UInt_t fNTrials;//number of trials with boost enabled
      Bool_t fRules;//A logical: should the tree be decomposed into a rule-based model?
 
      //Control options see C5.0Control
      Bool_t fControlSubset; //A logical: should the model evaluate groups of discrete predictors for splits?
      UInt_t fControlBands;
      Bool_t fControlWinnow;// A logical: should predictor winnowing (i.e feature selection) be used?
      Bool_t fControlNoGlobalPruning; //A logical to toggle whether the final, global pruning step to simplify the tree.
      Double_t fControlCF; //A number in (0, 1) for the confidence factor.
      UInt_t fControlMinCases;//an integer for the smallest number of samples that must be put in at least two of the splits.
      Bool_t fControlFuzzyThreshold;//A logical toggle to evaluate possible advanced splits of the data. See Quinlan (1993) for details and examples.
      Double_t fControlSample;//A value between (0, .999) that specifies the random proportion of the data should be used to train the model.
      Int_t fControlSeed;//An integer for the random number seed within the C code.
      Bool_t fControlEarlyStopping;// logical to toggle whether the internal method for stopping boosting should be used.
 
      UInt_t fMvaCounter;
      static Bool_t IsModuleLoaded;
 
      ROOT::R::TRFunctionImport predict;
      ROOT::R::TRFunctionImport C50;
      ROOT::R::TRFunctionImport C50Control;
      ROOT::R::TRFunctionImport asfactor;
      ROOT::R::TRObject *fModel;
      ROOT::R::TRObject fModelControl;
      std::vector <TString > ListOfVariables;
 
 
      // get help message text
      void GetHelpMessage() const override;
 
      ClassDefOverride(MethodC50, 0)
   };
} // namespace TMVA
#endif