Logo ROOT   6.30.04
Reference Guide
 All Namespaces Files Pages
probas.cxx
Go to the documentation of this file.
1 #include "TCanvas.h"
2 #include "TFile.h"
3 #include "TH2F.h"
4 #include "TIterator.h"
5 #include "TKey.h"
6 #include "TLegend.h"
7 #include "TList.h"
8 #include "TMVA/probas.h"
9 #include "TMVA/tmvaglob.h"
10 #include "TString.h"
11 
12 #include <iostream>
13 
14 using std::cout;
15 using std::endl;
16 
17 // this macro plots the MVA probability distributions (Signal and
18 // Background overlayed) of different MVA methods run in TMVA
19 // (e.g. running TMVAnalysis.C).
20 
21 // input: - Input file (result from TMVA)
22 // - use of TMVA plotting TStyle
23 void TMVA::probas(TString dataset, TString fin , Bool_t useTMVAStyle )
24 {
25  // set style and remove existing canvas'
26  TMVAGlob::Initialize( useTMVAStyle );
27 
28  // switches
29  const Bool_t Draw_CFANN_Logy = kFALSE;
30  const Bool_t Save_Images = kTRUE;
31 
32  // checks if file with name "fin" is already open, and if not opens one
33  TFile* file = TMVAGlob::OpenFile( fin );
34 
35  const Int_t width = 600; // size of canvas
36 
37  // this defines how many canvases we need
38  TCanvas *c = 0;
39 
40  // counter variables
41  Int_t countCanvas = 0;
42 
43  // list of existing MVAs
44  //const Int_t nveto = 1;
45  TString suffixSig = "_tr_S";
46  TString suffixBgd = "_tr_B";
47 
48  // search for the right histograms in full list of keys
49  TList methods;
50  UInt_t nmethods = TMVAGlob::GetListOfMethods( methods,file->GetDirectory(dataset.Data()) );
51  if (nmethods==0) {
52  cout << "--- Probas.C: no methods found!" << endl;
53  return;
54  }
55  TIter next(&methods);
56  TKey *key, *hkey;
57  char fname[200];
58  TH1* sig(0);
59  TH1* bgd(0);
60 
61 
62  while ( (key = (TKey*)next()) ) {
63  TDirectory * mDir = (TDirectory*)key->ReadObj();
64  TList titles;
65  UInt_t ni = TMVAGlob::GetListOfTitles( mDir, titles );
66  TString methodName;
67  TMVAGlob::GetMethodName(methodName,key);
68  if (ni==0) {
69  cout << "+++ No titles found for classifier: " << methodName << endl;
70  return;
71  }
72  TIter nextTitle(&titles);
73  TKey *instkey;
74  TDirectory *instDir;
75 
76  // iterate over all classifiers
77  while ( (instkey = (TKey *)nextTitle()) ) {
78  instDir = (TDirectory *)instkey->ReadObj();
79  TString instName = instkey->GetName();
80  TList h1hists;
81  UInt_t nhists = TMVAGlob::GetListOfKeys( h1hists, "TH1", instDir );
82  if (nhists==0) cout << "*** No histograms found!" << endl;
83  TIter nextInDir(&h1hists);
84  TString methodTitle;
85  TMVAGlob::GetMethodTitle(methodTitle,instDir);
86  Bool_t found = kFALSE;
87  while ( (hkey = (TKey*)nextInDir()) ) {
88  TH1 *th1 = (TH1*)hkey->ReadObj();
89  TString hname= th1->GetName();
90  if (hname.Contains( suffixSig ) && !hname.Contains( "Cut") &&
91  !hname.Contains("original") && !hname.Contains("smoothed")) {
92  // retrieve corresponding signal and background histograms
93  TString hnameS = hname;
94  TString hnameB = hname; hnameB.ReplaceAll("_S","_B");
95 
96  sig = (TH1*)instDir->Get( hnameS );
97  bgd = (TH1*)instDir->Get( hnameB );
98 
99  if (sig == 0 || bgd == 0) {
100  cout << "*** probas.C: big troubles in probas.... histogram: " << hname << " not found" << endl;
101  return;
102  }
103 
104  TH1* sigF(0);
105  TH1* bkgF(0);
106 
107  for (int i=0; i<= 5; i++) {
108  TString hspline = hnameS + Form("_smoothed_hist_from_spline%i",i);
109  sigF = (TH1*)instDir->Get( hspline );
110 
111  if (sigF) {
112  bkgF = (TH1*)instDir->Get( hspline.ReplaceAll("_tr_S","_tr_B") );
113  break;
114  }
115  }
116  if (!sigF){
117  TString hspline = hnameS + TString("_smoothed_hist_from_KDE");
118  sigF = (TH1*)instDir->Get( hspline );
119 
120  if (sigF) {
121  bkgF = (TH1*)instDir->Get( hspline.ReplaceAll("_tr_S","_tr_B") );
122  }
123  }
124 
125  if ((sigF == NULL || bkgF == NULL) &&!hname.Contains("hist") ) {
126  cout << "*** probas.C: big troubles - did not find probability histograms" << endl;
127  return;
128  }
129  else {
130  // remove the signal suffix
131 
132  // check that exist
133  if (NULL != sigF && NULL != bkgF && NULL!=sig && NULL!=bgd) {
134 
135  found = kTRUE;
136  // chop off useless stuff
137  sig->SetTitle( TString("TMVA output for classifier: ") + methodTitle );
138 
139  // create new canvas
140  cout << "--- Book canvas no: " << countCanvas << endl;
141  char cn[20];
142  sprintf( cn, "canvas%d", countCanvas+1 );
143  c = new TCanvas( cn, Form("TMVA Output Fit Variables %s",methodTitle.Data()),
144  countCanvas*50+200, countCanvas*20, width, width*0.78 );
145 
146  // set the histogram style
147  TMVAGlob::SetSignalAndBackgroundStyle( sig, bgd );
148  TMVAGlob::SetSignalAndBackgroundStyle( sigF, bkgF );
149 
150  // frame limits (choose judicuous x range)
151  Float_t nrms = 4;
152  Float_t xmin = TMath::Max( TMath::Min(sig->GetMean() - nrms*sig->GetRMS(),
153  bgd->GetMean() - nrms*bgd->GetRMS() ),
154  sig->GetXaxis()->GetXmin() );
155  Float_t xmax = TMath::Min( TMath::Max(sig->GetMean() + nrms*sig->GetRMS(),
156  bgd->GetMean() + nrms*bgd->GetRMS() ),
157  sig->GetXaxis()->GetXmax() );
158  Float_t ymin = 0;
159  Float_t ymax = TMath::Max( sig->GetMaximum(), bgd->GetMaximum() )*1.5;
160 
161  if (Draw_CFANN_Logy && methodName == "CFANN") ymin = 0.01;
162 
163  // build a frame
164  Int_t nb = 500;
165  TH2F* frame = new TH2F( TString("frame") + sig->GetName() + "_proba", sig->GetTitle(),
166  nb, xmin, xmax, nb, ymin, ymax );
167  frame->GetXaxis()->SetTitle(methodTitle);
168  frame->GetYaxis()->SetTitle("Normalized");
169  TMVAGlob::SetFrameStyle( frame );
170 
171  // eventually: draw the frame
172  frame->Draw();
173 
174  if (Draw_CFANN_Logy && methodName == "CFANN") c->SetLogy();
175 
176  // overlay signal and background histograms
177  sig->SetMarkerColor( TMVAGlob::getSignalLine() );
178  sig->SetMarkerSize( 0.7 );
179  sig->SetMarkerStyle( 20 );
180  sig->SetLineWidth(1);
181 
182  bgd->SetMarkerColor( TMVAGlob::getBackgroundLine() );
183  bgd->SetMarkerSize( 0.7 );
184  bgd->SetMarkerStyle( 24 );
185  bgd->SetLineWidth(1);
186 
187  sig->Draw("samee");
188  bgd->Draw("samee");
189 
190  sigF->SetFillStyle( 0 );
191  bkgF->SetFillStyle( 0 );
192  sigF->Draw("samehist");
193  bkgF->Draw("samehist");
194 
195  // redraw axes
196  frame->Draw("sameaxis");
197 
198  // Draw legend
199  TLegend *legend= new TLegend( c->GetLeftMargin(), 1 - c->GetTopMargin() - 0.2,
200  c->GetLeftMargin() + 0.4, 1 - c->GetTopMargin() );
201  legend->AddEntry(sig,"Signal data","P");
202  legend->AddEntry(sigF,"Signal PDF","L");
203  legend->AddEntry(bgd,"Background data","P");
204  legend->AddEntry(bkgF,"Background PDF","L");
205  legend->Draw("same");
206  legend->SetBorderSize(1);
207  legend->SetMargin( 0.3 );
208 
209  // save canvas to file
210  c->Update();
211  TMVAGlob::plot_logo();
212  sprintf( fname, "%s/plots/mva_pdf_%s_c%i",dataset.Data(), methodTitle.Data(), countCanvas+1 );
213  if (Save_Images) TMVAGlob::imgconv( c, fname );
214  countCanvas++;
215  }
216  }
217  }
218 
219  }
220  if(!found){
221  cout << "--- No PDFs found for method " << methodTitle << ". Did you request \"CreateMVAPdfs\" in the option string?" << endl;
222  }
223  }
224  }
225 }