| 1 | #ifndef MARS_MRanForestCalc
|
|---|
| 2 | #define MARS_MRanForestCalc
|
|---|
| 3 |
|
|---|
| 4 | #ifndef MARS_MTask
|
|---|
| 5 | #include "MTask.h"
|
|---|
| 6 | #endif
|
|---|
| 7 |
|
|---|
| 8 | #ifndef ROOT_TObjArray
|
|---|
| 9 | #include <TObjArray.h>
|
|---|
| 10 | #endif
|
|---|
| 11 |
|
|---|
| 12 | #ifndef ROOT_TArrayD
|
|---|
| 13 | #include <TArrayD.h>
|
|---|
| 14 | #endif
|
|---|
| 15 |
|
|---|
| 16 | class MDataArray;
|
|---|
| 17 | class MParameterD;
|
|---|
| 18 | class MHMatrix;
|
|---|
| 19 |
|
|---|
| 20 | class MRanForestCalc : public MTask
|
|---|
| 21 | {
|
|---|
| 22 | public:
|
|---|
| 23 | enum EstimationMode_t
|
|---|
| 24 | {
|
|---|
| 25 | kMean,
|
|---|
| 26 | kMaximum,
|
|---|
| 27 | kFit
|
|---|
| 28 | };
|
|---|
| 29 |
|
|---|
| 30 | private:
|
|---|
| 31 | static const TString gsDefName; //! Default Name
|
|---|
| 32 | static const TString gsDefTitle; //! Default Title
|
|---|
| 33 | static const TString gsNameOutput; //! Default Output name
|
|---|
| 34 |
|
|---|
| 35 | MDataArray *fData; //! Used to store the MDataChains to get the event values
|
|---|
| 36 | MParameterD *fRFOut; //! Used to store result
|
|---|
| 37 | MHMatrix *fTestMatrix; //! Test Matrix used in Process (together with MMatrixLoop)
|
|---|
| 38 |
|
|---|
| 39 | TObjArray fEForests; //! List of forests read or to be written
|
|---|
| 40 |
|
|---|
| 41 | Int_t fNumTrees; //! Training parameters
|
|---|
| 42 | Int_t fNumTry; //! Training parameters
|
|---|
| 43 | Int_t fNdSize; //! Training parameters
|
|---|
| 44 |
|
|---|
| 45 | Int_t fNumObsoleteVariables; //! Training parameters
|
|---|
| 46 | Bool_t fLastDataColumnHasWeights; //! Training parameters
|
|---|
| 47 |
|
|---|
| 48 | TString fFileName; // File name to forest
|
|---|
| 49 | TString fNameOutput; // Name of output container
|
|---|
| 50 |
|
|---|
| 51 | Bool_t fDebug; // Debugging of eventloop while training on/off
|
|---|
| 52 |
|
|---|
| 53 | EstimationMode_t fEstimationMode; // Mode of estimation in case of multi random forest regression
|
|---|
| 54 |
|
|---|
| 55 | private:
|
|---|
| 56 | // MTask
|
|---|
| 57 | Int_t PreProcess(MParList *plist);
|
|---|
| 58 | Int_t Process();
|
|---|
| 59 |
|
|---|
| 60 | // MRanForestCalc
|
|---|
| 61 | Int_t ReadForests(MParList &plist);
|
|---|
| 62 |
|
|---|
| 63 | // MParContainer
|
|---|
| 64 | Int_t ReadEnv(const TEnv &env, TString prefix, Bool_t print);
|
|---|
| 65 |
|
|---|
| 66 | // Train Interface
|
|---|
| 67 | Int_t Train(const MHMatrix &n, const TArrayD &grid, Int_t ver);
|
|---|
| 68 |
|
|---|
| 69 | public:
|
|---|
| 70 | MRanForestCalc(const char *name=NULL, const char *title=NULL);
|
|---|
| 71 | ~MRanForestCalc();
|
|---|
| 72 |
|
|---|
| 73 | // TObject
|
|---|
| 74 | void Print(Option_t *o="") const; //*MENU*
|
|---|
| 75 |
|
|---|
| 76 | // Setter for estimation
|
|---|
| 77 | void SetFileName(TString filename) { fFileName = filename; }
|
|---|
| 78 | void SetEstimationMode(EstimationMode_t op) { fEstimationMode = op; }
|
|---|
| 79 | void SetNameOutput(TString name=gsNameOutput) { fNameOutput = name; }
|
|---|
| 80 |
|
|---|
| 81 | // Setter for training
|
|---|
| 82 | void SetNumTrees(UShort_t n=100) { fNumTrees = n; }
|
|---|
| 83 | void SetNdSize(UShort_t n=5) { fNdSize = n; }
|
|---|
| 84 | void SetNumTry(UShort_t n=0) { fNumTry = n; }
|
|---|
| 85 | void SetDebug(Bool_t b=kTRUE) { fDebug = b; }
|
|---|
| 86 |
|
|---|
| 87 | void SetNumObsoleteVariables(Int_t n=1) { fNumObsoleteVariables = n; }
|
|---|
| 88 | void SetLastDataColumnHasWeights(Bool_t b=kTRUE) { fLastDataColumnHasWeights = b; }
|
|---|
| 89 |
|
|---|
| 90 | // Train Interface
|
|---|
| 91 | Int_t TrainMultiRF(const MHMatrix &n, const TArrayD &grid)
|
|---|
| 92 | {
|
|---|
| 93 | // One yes/no-classification forest is trained for each bin
|
|---|
| 94 | return Train(n, grid, 0);
|
|---|
| 95 | }
|
|---|
| 96 | Int_t TrainSingleRF(const MHMatrix &n, const TArrayD &grid=TArrayD())
|
|---|
| 97 | {
|
|---|
| 98 | // w/o Grid: Last Column contains classifier
|
|---|
| 99 | // w/ Grid: Last Column will be converted by grid into classifier
|
|---|
| 100 | return Train(n, grid, grid.GetSize()==0 ? 2 : 1);
|
|---|
| 101 | }
|
|---|
| 102 | Int_t TrainRegression(const MHMatrix &n)
|
|---|
| 103 | {
|
|---|
| 104 | // Use last column for regression
|
|---|
| 105 | return Train(n, TArrayD(), 3);
|
|---|
| 106 | }
|
|---|
| 107 |
|
|---|
| 108 | // Test Interface
|
|---|
| 109 | void SetTestMatrix(MHMatrix *m=0) { fTestMatrix=m; }
|
|---|
| 110 | void InitMapping(MHMatrix *m=0) { fTestMatrix=m; }
|
|---|
| 111 |
|
|---|
| 112 | ClassDef(MRanForestCalc, 1) // Task to calculate RF output and for RF training
|
|---|
| 113 | };
|
|---|
| 114 |
|
|---|
| 115 | #endif
|
|---|
