Logo ROOT   6.10/09
Reference Guide
variablesMultiClass.cxx
Go to the documentation of this file.
2 
3 
4 // this macro plots the distributions of the different input variables
5 // used in TMVA (e.g. running TMVAnalysis.C). Signal and Background are overlayed.
6 
7 // input: - Input file (result from TMVA),
8 // - normal/decorrelated/PCA
9 // - use of TMVA plotting TStyle
10 void TMVA::variablesMultiClass(TString dataset, TString fin , TString dirName , TString title,
11  Bool_t /* isRegression */, Bool_t useTMVAStyle )
12 {
13  TString outfname = dirName;
14  TString tmp = dirName;
15  tmp.ReplaceAll("InputVariables_","");
16  outfname.ToLower(); outfname.ReplaceAll( "input", "" );
17 
18  // set style and remove existing canvas'
19  TMVAGlob::Initialize( useTMVAStyle );
20 
21  // obtain shorter histogram title
22  TString htitle = title;
23  htitle.ReplaceAll("variables ","variable");
24  htitle.ReplaceAll("and target(s)","");
25  htitle.ReplaceAll("(training sample)","");
26 
27  // checks if file with name "fin" is already open, and if not opens one
28  TFile* file = TMVAGlob::OpenFile( fin );
29 
30  TDirectory* dir = (TDirectory*)file->GetDirectory(dataset.Data())->Get( dirName );
31  if (dir==0) {
32  cout << "No information about " << title << " available in directory " << dirName << " of file " << fin << endl;
33  return;
34  }
35  dir->cd();
36 
37  // how many plots are in the directory?
39 
40  // define Canvas layout here!
41  // default setting
42  Int_t xPad; // no of plots in x
43  Int_t yPad; // no of plots in y
44  Int_t width; // size of canvas
45  Int_t height;
46  switch (noPlots) {
47  case 1:
48  xPad = 1; yPad = 1; width = 550; height = 0.90*width; break;
49  case 2:
50  xPad = 2; yPad = 1; width = 600; height = 0.50*width; break;
51  case 3:
52  xPad = 3; yPad = 1; width = 900; height = 0.4*width; break;
53  case 4:
54  xPad = 2; yPad = 2; width = 600; height = width; break;
55  default:
56  xPad = 3; yPad = 2; width = 800; height = 0.55*width; break;
57  }
58 
59  Int_t noPadPerCanv = xPad * yPad ;
60 
61  // counter variables
62  Int_t countCanvas = 0;
63  Int_t countPad = 0;
64 
65  // loop over all objects in directory
66  TCanvas* canv = 0;
67  Bool_t createNewFig = kFALSE;
68  TIter next(dir->GetListOfKeys());
69 
70  std::vector<TString> varnames(TMVAGlob::GetInputVariableNames(dir));
71  std::vector<TString> classnames(TMVAGlob::GetClassNames(dir));
72 
73 
74  std::vector<TString>::iterator variter = varnames.begin();
75  std::vector<TString>::iterator classiter = classnames.begin();
76 
77  /*
78  std::vector<TString>::const_iterator variter = varnames.begin();
79  std::cout << "Available variables:" << std::endl;
80  while(variter != varnames.end()){
81  std::cout << *variter << std::endl;
82  variter++;
83  }
84 
85  std::vector<TString>::const_iterator classiter = classnames.begin();
86  std::cout << "Available classes:" << std::endl;
87  while(classiter != classnames.end()){
88  std::cout << *classiter << std::endl;
89  classiter++;
90  }
91  */
92 
93  variter = varnames.begin();
94  for(; variter!=varnames.end(); ++variter){
95 
96  //create new canvas
97  if (countPad%noPadPerCanv==0) {
98  ++countCanvas;
99  canv = new TCanvas( Form("canvas%d", countCanvas), title,
100  countCanvas*50+50, countCanvas*20, width, height );
101  canv->Divide(xPad,yPad);
102  canv->Draw();
103  }
104  TPad* cPad = (TPad*)canv->cd(countPad++%noPadPerCanv+1);
105  classiter = classnames.begin();
106 
107  TObjArray hists;
108  for(; classiter!=classnames.end(); ++classiter){
109  //assemble histogram names
110  TString hname(*variter + "__" + *classiter + "_" + tmp);
111  TH1 *hist = (TH1*)dir->Get(hname);
112  //cout << "Looking for histgram " << hname << endl;
113  if (hist == NULL) {
114  cout << "ERROR!!! couldn't find " << *variter << " histogram for class " << *classiter << endl;
115  //exit(1);
116  return;
117  }
118  hists.Add(hist);
119  }
120 
121  // this is set but not stored during plot creation in MVA_Factory
122  //TMVAGlob::SetSignalAndBackgroundStyle(((TH1*)hists[0]), ((TH1*)hists[1]));
123  TMVAGlob::SetMultiClassStyle( &hists );
124 
125  ((TH1*)hists.First())->SetTitle( TString( htitle ) + ": " + *variter );
126 
127  TMVAGlob::SetFrameStyle( ((TH1*)hists.First()), 1.2 );
128 
129  // normalise all histograms and find maximum
130  Float_t histmax = -1;
131  for(Int_t i=0; i<hists.GetEntriesFast(); ++i){
132  TMVAGlob::NormalizeHist((TH1*)hists[i] );
133  if(((TH1*)hists[i])->GetMaximum() > histmax)
134  histmax = ((TH1*)hists[i])->GetMaximum();
135  }
136 
137  // finally plot and overlay
138  Float_t sc = 1.1;
139  if (countPad == 1) sc = 1.3;
140  ((TH1*)hists.First())->SetMaximum( histmax*sc );
141 
142  ((TH1*)hists.First())->Draw( "hist" );
143  cPad->SetLeftMargin( 0.17 );
144  ((TH1*)hists.First())->GetYaxis()->SetTitleOffset( 1.70 );
145 
146  for(Int_t i=1; i<hists.GetEntriesFast(); ++i){
147 
148  ((TH1*)hists[i])->Draw("histsame");
149  TString ytit = TString("(1/N) ") + ((TH1*)hists[i])->GetYaxis()->GetTitle();
150  ((TH1*)hists[i])->GetYaxis()->SetTitle( ytit ); // histograms are normalised
151 
152  }
153 
154  // Draw legend
155  if (countPad == 1) {
156  TLegend *legend= new TLegend( cPad->GetLeftMargin(),
157  1-cPad->GetTopMargin()-.15,
158  cPad->GetLeftMargin()+.4,
159  1-cPad->GetTopMargin() );
160  legend->SetFillStyle(1);
161 
162  classiter = classnames.begin();
163 
164  for(Int_t i=0; i<hists.GetEntriesFast(); ++i, ++classiter){
165  legend->AddEntry(((TH1*)hists[i]),*classiter,"F");
166  }
167 
168  legend->SetBorderSize(1);
169  legend->SetMargin( 0.3 );
170  legend->Draw("same");
171  }
172 
173  // redraw axes
174  ((TH1*)hists.First())->Draw("sameaxis");
175 
176 
177  // text for overflows
178  Int_t nbin = ((TH1*)hists.First())->GetNbinsX();
179  Double_t dxu = ((TH1*)hists.First())->GetBinWidth(0);
180  Double_t dxo = ((TH1*)hists.First())->GetBinWidth(nbin+1);
181  TString uoflow = "";
182 
183  classiter = classnames.begin();
184  for(Int_t i=0; i<hists.GetEntriesFast(); ++i, ++classiter){
185  if(((TH1*)hists[i])->GetBinContent(0)!=0 || ((TH1*)hists[i])->GetBinContent(nbin+1)!=0){
186  uoflow += *classiter;
187  uoflow += Form( " U/O-flow: %.1f / %.1f %%",
188  ((TH1*)hists[i])->GetBinContent(0)*dxu*100, ((TH1*)hists[i])->GetBinContent(nbin+1)*dxo*100);
189  }
190  }
191 
192  TText* t = new TText( 0.98, 0.14, uoflow );
193  t->SetNDC();
194  t->SetTextSize( 0.040 );
195  t->SetTextAngle( 90 );
196  t->AppendPad();
197 
198 
199  // save canvas to file
200  if (countPad%noPadPerCanv==0) {
201  TString fname = dataset+Form( "/plots/%s_c%i", outfname.Data(), countCanvas );
203  TMVAGlob::imgconv( canv, fname );
204  createNewFig = kFALSE;
205  }
206  else {
207  createNewFig = kTRUE;
208  }
209  }
210 
211  if (createNewFig) {
212  TString fname = dataset+Form( "/plots/%s_c%i", outfname.Data(), countCanvas );
214  TMVAGlob::imgconv( canv, fname );
215  createNewFig = kFALSE;
216  }
217 
218  return;
219 }
void variablesMultiClass(TString dataset, TString fin="TMVA.root", TString dirName="InputVariables_Id", TString title="TMVA Input Variables", Bool_t isRegression=kFALSE, Bool_t useTMVAStyle=kTRUE)
An array of TObjects.
Definition: TObjArray.h:37
void imgconv(TCanvas *c, const TString &fname)
Definition: tmvaglob.cxx:212
Float_t GetLeftMargin() const
Definition: TAttPad.h:44
virtual TList * GetListOfKeys() const
Definition: TDirectory.h:148
std::vector< TString > GetInputVariableNames(TDirectory *dir)
Definition: tmvaglob.cxx:429
This class displays a legend box (TPaveText) containing several legend entries.
Definition: TLegend.h:23
Int_t GetNumberOfInputVariables(TDirectory *dir)
Definition: tmvaglob.cxx:413
virtual TObject * Get(const char *namecycle)
Return pointer to object identified by namecycle.
Definition: TDirectory.cxx:729
float Float_t
Definition: RtypesCore.h:53
TString & ReplaceAll(const TString &s1, const TString &s2)
Definition: TString.h:640
TFile * OpenFile(const TString &fin)
Definition: tmvaglob.cxx:192
virtual void Draw(Option_t *option="")
Draw this legend with its current attributes.
Definition: TLegend.cxx:452
std::vector< TString > GetClassNames(TDirectory *dir)
Definition: tmvaglob.cxx:462
A ROOT file is a suite of consecutive data records (TKey instances) with a well defined format...
Definition: TFile.h:46
TVirtualPad * cd(Int_t subpadnumber=0)
Set current canvas & pad.
Definition: TCanvas.cxx:679
void NormalizeHist(TH1 *h)
Definition: tmvaglob.cxx:308
void SetFrameStyle(TH1 *frame, Float_t scale=1.0)
Definition: tmvaglob.cxx:77
Basic string class.
Definition: TString.h:129
void ToLower()
Change string to lower-case.
Definition: TString.cxx:1099
int Int_t
Definition: RtypesCore.h:41
bool Bool_t
Definition: RtypesCore.h:59
void SetMargin(Float_t margin)
Definition: TLegend.h:69
virtual void SetFillStyle(Style_t fstyle)
Set the fill area style.
Definition: TAttFill.h:39
#define NULL
Definition: RtypesCore.h:88
TLegend * legend
Definition: pirndm.C:35
virtual void AppendPad(Option_t *option="")
Append graphics object to current pad.
Definition: TObject.cxx:112
Base class for several text objects.
Definition: TText.h:23
virtual void SetNDC(Bool_t isNDC=kTRUE)
Set NDC mode on if isNDC = kTRUE, off otherwise.
Definition: TText.cxx:814
void SetMultiClassStyle(TObjArray *hists)
Definition: tmvaglob.cxx:47
TObject * First() const
Return the object in the first slot.
Definition: TObjArray.cxx:471
th1 Draw()
void Initialize(Bool_t useTMVAStyle=kTRUE)
Definition: tmvaglob.cxx:176
virtual void SetTextAngle(Float_t tangle=0)
Set the text angle.
Definition: TAttText.h:42
The most important graphics class in the ROOT system.
Definition: TPad.h:29
Int_t GetEntriesFast() const
Definition: TObjArray.h:64
char * Form(const char *fmt,...)
void plot_logo(Float_t v_scale=1.0, Float_t skew=1.0)
Definition: tmvaglob.cxx:263
virtual TDirectory * GetDirectory(const char *apath, Bool_t printError=false, const char *funcname="GetDirectory")
Find a directory named "apath".
const Bool_t kFALSE
Definition: RtypesCore.h:92
The Canvas class.
Definition: TCanvas.h:31
double Double_t
Definition: RtypesCore.h:55
TLegendEntry * AddEntry(const TObject *obj, const char *label="", Option_t *option="lpf")
Add a new entry to this legend.
Definition: TLegend.cxx:359
Describe directory structure in memory.
Definition: TDirectory.h:34
The TH1 histogram class.
Definition: TH1.h:56
virtual void Draw(Option_t *option="")
Draw a canvas.
Definition: TCanvas.cxx:817
Float_t GetTopMargin() const
Definition: TAttPad.h:46
virtual Bool_t cd(const char *path=0)
Change current directory to "this" directory.
Definition: TDirectory.cxx:435
virtual void Divide(Int_t nx=1, Int_t ny=1, Float_t xmargin=0.01, Float_t ymargin=0.01, Int_t color=0)
Automatic pad generation by division.
Definition: TPad.cxx:1135
Definition: file.py:1
void Add(TObject *obj)
Definition: TObjArray.h:73
virtual void SetTextSize(Float_t tsize=1)
Set the text size.
Definition: TAttText.h:46
const Bool_t kTRUE
Definition: RtypesCore.h:91
virtual void SetBorderSize(Int_t bordersize=4)
Definition: TPave.h:70
virtual void SetLeftMargin(Float_t leftmargin)
Set Pad left margin in fraction of the pad width.
Definition: TAttPad.cxx:110
const char * Data() const
Definition: TString.h:347