| 1 |
nmohr |
1.1 |
import sys,re,ROOT
|
| 2 |
|
|
ROOT.gROOT.SetBatch(True)
|
| 3 |
|
|
from sample_parser import ParseInfo
|
| 4 |
|
|
from TreeCache import TreeCache
|
| 5 |
|
|
|
| 6 |
|
|
class RegressionTrainer():
|
| 7 |
|
|
def __init__(self, config):
|
| 8 |
|
|
vhbb_name_space = config.get('VHbbNameSpace','library')
|
| 9 |
|
|
ROOT.gSystem.Load(vhbb_name_space)
|
| 10 |
|
|
|
| 11 |
|
|
self.__weight = config.get("TrainRegression","weight")
|
| 12 |
|
|
self.__vars = config.get("TrainRegression","vars").split()
|
| 13 |
|
|
self.__target = config.get("TrainRegression","target")
|
| 14 |
|
|
self.__cut = config.get("TrainRegression","cut")
|
| 15 |
|
|
self.__title = config.get("TrainRegression","name")
|
| 16 |
|
|
self.__signals = config.get("TrainRegression","signals")
|
| 17 |
|
|
self.__regOptions = config.get("TrainRegression","options")
|
| 18 |
|
|
path = config.get('Directories','PREPout')
|
| 19 |
|
|
samplesinfo=config.get('Directories','samplesinfo')
|
| 20 |
|
|
self.__info = ParseInfo(samplesinfo,path)
|
| 21 |
|
|
self.__samples = self.__info.get_samples(self.__signals)
|
| 22 |
|
|
self.__tc = TreeCache([self.__cut],self.__samples,path,config)
|
| 23 |
|
|
self.__trainCut = config.get("TrainRegression","trainCut")
|
| 24 |
|
|
self.__testCut = config.get("TrainRegression","testCut")
|
| 25 |
|
|
self.__config = config
|
| 26 |
|
|
|
| 27 |
|
|
|
| 28 |
|
|
def train(self):
|
| 29 |
|
|
signals = []
|
| 30 |
|
|
signalsTest = []
|
| 31 |
|
|
for job in self.__samples:
|
| 32 |
|
|
print '\tREADING IN %s AS SIG'%job.name
|
| 33 |
|
|
signals.append(self.__tc.get_tree(job,'%s & %s' %(self.__cut,self.__trainCut)))
|
| 34 |
|
|
signalsTest.append(self.__tc.get_tree(job,'%s & %s'%(self.__cut,self.__testCut)))
|
| 35 |
|
|
|
| 36 |
|
|
sWeight = 1.
|
| 37 |
|
|
fnameOutput='training_Reg_%s.root' %(self.__title)
|
| 38 |
|
|
output = ROOT.TFile.Open(fnameOutput, "RECREATE")
|
| 39 |
|
|
|
| 40 |
nmohr |
1.2 |
factory = ROOT.TMVA.Factory('MVA', output, '!V:!Silent:!Color:!DrawProgressBar:Transformations=I:AnalysisType=Regression')
|
| 41 |
nmohr |
1.1 |
#factory.SetSignalWeightExpression( self.__weight )
|
| 42 |
|
|
#set input trees
|
| 43 |
|
|
for i, signal in enumerate(signals):
|
| 44 |
|
|
factory.AddRegressionTree( signal, sWeight, ROOT.TMVA.Types.kTraining)
|
| 45 |
|
|
factory.AddRegressionTree( signalsTest[i], sWeight, ROOT.TMVA.Types.kTesting)
|
| 46 |
|
|
self.__apply = []
|
| 47 |
|
|
p = re.compile(r'hJet_\w+')
|
| 48 |
|
|
for var in self.__vars:
|
| 49 |
|
|
factory.AddVariable(var,'D') # add the variables
|
| 50 |
|
|
self.__apply.append(p.sub(r'\g<0>[0]', var))
|
| 51 |
|
|
|
| 52 |
|
|
factory.AddTarget( self.__target )
|
| 53 |
|
|
mycut = ROOT.TCut( self.__cut )
|
| 54 |
|
|
factory.BookMethod(ROOT.TMVA.Types.kBDT,'BDT_REG_%s'%(self.__title),self.__regOptions) # book an MVA method
|
| 55 |
|
|
factory.TrainAllMethods()
|
| 56 |
|
|
factory.TestAllMethods()
|
| 57 |
|
|
factory.EvaluateAllMethods()
|
| 58 |
|
|
output.Write()
|
| 59 |
|
|
regDict = dict(zip(self.__vars, self.__apply))
|
| 60 |
|
|
self.__config.set('Regression', 'regWeight', '../data/MVA_BDT_REG_%s.weights.xml' %self.__title)
|
| 61 |
|
|
self.__config.set('Regression', 'regDict', '%s' %regDict)
|
| 62 |
|
|
self.__config.set('Regression', 'regVars', '%s' %self.__vars)
|
| 63 |
|
|
for section in self.__config.sections():
|
| 64 |
|
|
if not section == 'Regression':
|
| 65 |
|
|
self.__config.remove_section(section)
|
| 66 |
|
|
with open('8TeVconfig/appReg', 'w') as configfile:
|
| 67 |
|
|
self.__config.write(configfile)
|
| 68 |
|
|
with open('8TeVconfig/appReg', 'r') as configfile:
|
| 69 |
|
|
for line in configfile:
|
| 70 |
|
|
print line.strip()
|
| 71 |
|
|
|
