sig_dataDev=dataDev[dataDev.category==1]
bkg_dataDev=dataDev[dataDev.category==0]
sig_dataVal=dataVal[dataVal.category==1]
bkg_dataVal=dataVal[dataVal.category==0]
fomEvo = []
fomCut = []
bkgEff = []
sigEff = []
sig_Init = dataVal[dataVal.category == 1].weight.sum() * 35866 * 2
bkg_Init = dataVal[dataVal.category == 0].weight.sum() * 35866 * 2
for cut in np.arange(0.0, 0.9999, 0.001):
sig, bkg = getYields(dataVal, cut=cut, luminosity=luminosity)
if sig[0] > 0 and bkg[0] > 0:
fom, fomUnc = FullFOM(sig, bkg)
fomEvo.append(fom)
fomCut.append(cut)
bkgEff.append(bkg[0]/bkg_Init)
sigEff.append(sig[0]/sig_Init)
print "Plotting"
plt.figure(figsize=(7,6))
plt.hist(sig_dataDev["NN"], 50, facecolor='blue', alpha=0.7, normed=1, weights=sig_dataDev["weight"])
plt.hist(bkg_dataDev["NN"], 50, facecolor='red', alpha=0.7, normed=1, weights=bkg_dataDev["weight"])
plt.hist(sig_dataVal["NN"], 50, color='blue', alpha=1, normed=1, weights=sig_dataVal["weight"], histtype="step")
plt.hist(bkg_dataVal["NN"], 50, color='red', alpha=1, normed=1, weights=bkg_dataVal["weight"], histtype="step")
plt.xlabel('NN output')
scoreDev = model.evaluate(XDev, YDev, sample_weight=weightDev, verbose=0)
scoreVal = model.evaluate(XVal, YVal, sample_weight=weightVal, verbose=0)
print ""
cohen_kappa = cohen_kappa_score(YVal, valPredict.round())
print "Calculating parameters"
dataDev["NN"] = devPredict
dataVal["NN"] = valPredict
sig_dataDev = dataDev[dataDev.category == 1]
bkg_dataDev = dataDev[dataDev.category == 0]
sig_dataVal = dataVal[dataVal.category == 1]
bkg_dataVal = dataVal[dataVal.category == 0]
tmpSig, tmpBkg = getYields(dataVal)
sigYield, sigYieldUnc = tmpSig
bkgYield, bkgYieldUnc = tmpBkg
fomEvo = []
fomCut = []
bkgEff = []
sigEff = []
sig_Init = dataVal[dataVal.category == 1].weight.sum() * 35866 * 2
bkg_Init = dataVal[dataVal.category == 0].weight.sum() * 35866 * 2
for cut in np.arange(0.0, 0.9999, 0.001):
sig, bkg = getYields(dataVal, cut=cut, luminosity=luminosity)
if sig[0] > 0 and bkg[0] > 0:
devPredict = model.predict(XDev)
valPredict = model.predict(XVal)
#Getting scores
if args.verbose:
print("Getting scores")
scoreDev = model.evaluate(XDev, YDev, sample_weight=weightDev, verbose = 0)
scoreVal = model.evaluate(XVal, YVal, sample_weight=weightVal, verbose = 0)
# CAlculating FOM
if args.verbose:
print "Calculating FOM:"
dataVal["NN"] = valPredict
tmpSig, tmpBkg = getYields(dataVal)
sigYield, sigYieldUnc = tmpSig
bkgYield, bkgYieldUnc = tmpBkg
sigDataVal = dataVal[dataVal.category==1]
bkgDataVal = dataVal[dataVal.category==0]
fomEvo = []
fomCut = []
for cut in np.arange(0.0, 0.9999999, 0.001):
sig, bkg = getYields(dataVal, cut=cut)
if sig[0] > 0 and bkg[0] > 0:
fom, fomUnc = FullFOM(sig, bkg)
fomEvo.append(fom)
fomCut.append(cut)