def ROC_plot(SEff, BRej, model_id=""):
root.TMVA.Tools.Instance()
aplt.set_atlas_style()
area = 0
for index, xval in enumerate(SEff[:-1]):
delta = SEff[index + 1] - xval
area -= delta * BRej[index]
XPlot, YPlot = array("d"), array("d")
for x, y in zip(SEff, BRej):
XPlot.append(x)
YPlot.append(y)
ndots = len(SEff)
fig, ax = aplt.subplots(1, 1, name="fig", figsize=(800, 800))
curve = root.TGraph(ndots, XPlot, YPlot)
curve.SetLineColor(4)
curve.SetLineWidth(2)
curve.GetXaxis().SetTitle(f"BDTgrad_{model_id} signal efficiency")
curve.GetYaxis().SetTitle(f"BDTgrad_{model_id} background rejection")
curve.SetMarkerSize(0)
ax.plot(curve)
ax.text(0.19, 0.92, "#sqrt{s} = 13 TeV, 139 fb^{-1}", size=27, align=13)
ax.text(0.195, 0.84, f"Area under ROC-curve: {round(area, 3)}")
fig.savefig(f"BDTgrad_{model_id}_ROC_curve.pdf")
def main():
# Set the ATLAS Style
aplt.set_atlas_style()
# Create a figure and axes
fig, ax = aplt.subplots(1, 1, name="fig1", figsize=(800, 600))
# Randomly fill a 2D histogram
hist = root.TH2F("hist", "Random Histogram", 50, 0, 10, 40, 0, 200)
root.fillHist2D(hist, 50000)
# Draw the histogram on these axes
ax.plot2d(hist, "COLZ")
# Change pad margins to allow space for z-axis colour bar and for ATLAS label
ax.set_pad_margins(right=0.20, top=0.08)
# Set axis titles
ax.set_xlabel("X")
ax.set_ylabel("Y")
ax.set_zlabel("Events / (0.2 #times 5)", titleoffset=1.2)
# Add the ATLAS Label
aplt.atlas_label(ax.pad.GetLeftMargin(), 0.97, text="Internal", align=13)
ax.text(1 - ax.pad.GetRightMargin(),
0.97,
"#sqrt{s} = 13 TeV, 139 fb^{-1}",
size=22,
align=33)
# Save the plot as a PDF
fig.savefig("hist2D.pdf")
def BDT_output_hist_plot(SDataframe,
BDataframe,
model_id="",
output_name="output"):
root.TMVA.Tools.Instance()
aplt.set_atlas_style()
Min = np.max(
(np.min(SDataframe[output_name]), np.min(BDataframe[output_name])))
Max = np.min(
(np.max(SDataframe[output_name]), np.max(BDataframe[output_name])))
# Min =-math.ceil(abs(Min))
# Max = math.ceil(Max)
SHist = root.TH1F("", "", 50, Min, Max)
BHist = root.TH1F("", "", 50, Min, Max)
SHist.SetStats(False)
BHist.SetStats(False)
SHist.SetMinimum(0)
BHist.SetMinimum(0)
SHist.SetLineWidth(2)
BHist.SetLineWidth(2)
SHist.SetLineColor(4)
BHist.SetLineColor(2)
SHist.SetMarkerSize(0)
BHist.SetMarkerSize(0)
SHist.SetFillColorAlpha(4, 0.2)
BHist.SetFillColor(2)
BHist.SetFillStyle(3004)
for out, weight in zip(SDataframe[output_name],
SDataframe["weightModified"]):
SHist.Fill(out, weight)
for out, weight in zip(BDataframe[output_name],
BDataframe["weightModified"]):
BHist.Fill(out, weight)
hists_max = np.max((get_hist_max(SHist, 50), get_hist_max(BHist, 50)))
margins = [Min, Max, 0, hists_max]
fig, ax = aplt.subplots(1, 1, name="", figsize=(800, 600))
ax.plot(SHist, margins, "E1")
ax.plot(BHist, margins, "E1")
ax.add_margins(top=0.1)
ax.set_xlabel(f"BDTgrad_{model_id} classifier response")
ax.set_ylabel("Fraction of events")
ax.text(0.2, 0.92, "#sqrt{s} = 13 TeV, 139 fb^{-1}", size=27, align=13)
legend = root.TLegend(0.65, 0.8, 0.95, 0.92)
legend.SetFillColorAlpha(0, 0)
legend.AddEntry(SHist, "Signal", "F")
legend.AddEntry(BHist, "Background", "F")
legend.Draw()
fig.savefig(f"BDTgrad_{model_id}_output.pdf")
def main():
# Set the ATLAS Style
aplt.set_atlas_style()
# Create a figure and axes
fig, (ax1, ax2) = aplt.ratio_plot(name="fig1",
figsize=(800, 800),
hspace=0.05)
# Define a distribution
sqroot = root.TF1("sqroot", "x*gaus(0) + [3]*abs(sin(x)/x)", 0, 10)
sqroot.SetParameters(10, 4, 1, 20)
# Randomly fill two histograms according to the above distribution
hist1 = root.TH1F("hist1", "Random Histogram 1", 50, 0, 10)
hist1.FillRandom("sqroot", 20000)
sqroot.SetParameters(10, 4, 1.1, 20)
hist2 = root.TH1F("hist2", "Random Histogram 2", 50, 0, 10)
hist2.FillRandom("sqroot", 20000)
# Draw the histograms on these axes
ax1.plot(hist1, linecolor=root.kRed + 1, label="Red", labelfmt="L")
ax1.plot(hist2, linecolor=root.kBlue + 1, label="Blue", labelfmt="L")
# Draw line at y=1 in ratio panel
line = root.TLine(ax1.get_xlim()[0], 1, ax1.get_xlim()[1], 1)
ax2.plot(line)
# Calculate and draw the ratio
ratio_hist = hist1.Clone("ratio_hist")
ratio_hist.Divide(hist2)
ax2.plot(ratio_hist, linecolor=root.kBlack)
# Add extra space at top of plot to make room for labels
ax1.add_margins(top=0.16)
# Set axis titles
ax2.set_xlabel("X [GeV]")
ax1.set_ylabel("Events / 0.2 GeV")
ax2.set_ylabel("Red / Blue", loc="centre")
# Add extra space at top and bottom of ratio panel
ax2.add_margins(top=0.1, bottom=0.1)
# Go back to top axes to add labels
ax1.cd()
# Add the ATLAS Label
aplt.atlas_label(text="Internal", loc="upper left")
ax1.text(0.2, 0.84, "#sqrt{s} = 13 TeV, 139 fb^{-1}", size=22, align=13)
# Add legend
ax1.legend(loc=(0.78, 0.78, 1, 0.90))
# Save the plot as a PDF
fig.savefig("ratio.pdf")
def main():
# Set the ATLAS Style
aplt.set_atlas_style()
# Create a figure and axes
fig, ax = aplt.subplots(1, 1, name="fig1", figsize=(800, 600))
# Plot a line with Gaussian noise
x = np.arange(20)
y = 2 * x + np.random.normal(size=20)
yerr = np.random.normal(loc=np.linspace(1, 2, num=20), scale=0.1, size=20)
graph = ax.graph(x, y, yerr=yerr, label="Data", labelfmt="EP")
# Fit the graph; store graphics func but do not draw
graph.Fit("pol1", "0")
# Add extra space at top of plot to make room for labels
ax.add_margins(top=0.18, left=0.05, right=0.05, bottom=0.05)
# Plot fit function and extend its range to fill plot area
func = graph.GetFunction("pol1")
func.SetRange(*ax.get_xlim())
func.SetNpx(1000)
ax.plot(func,
linecolor=root.kRed + 1,
expand=False,
label="Fit",
labelfmt="L")
# Set axis titles
ax.set_xlabel("x")
ax.set_ylabel("y")
# Add the ATLAS Label
aplt.atlas_label(text="Simulation Internal", loc="upper left")
# Add legend
ax.legend(loc=(0.65, 0.8, 0.95, 0.92))
# Save the plot as a PDF
fig.savefig("numpy_graph.pdf")
def main():
# Set the ATLAS Style
aplt.set_atlas_style()
# Create a figure and axes
fig, ax = aplt.subplots(1, 1, name="fig1", figsize=(800, 600))
# Define a distribution
sqroot = root.TF1("sqroot", "x*gaus(0) + [3]*abs(sin(x)/x)", 0, 10)
sqroot.SetParameters(10, 4, 1, 20)
# Randomly fill the histogram according to the above distribution
hist = root.TH1F("hist", "Random Histogram", 50, 0, 10)
hist.FillRandom("sqroot", 20000)
# Fit the histogram with the original distribution; store graphics func but do not draw
hist.Fit("sqroot", "0")
# Draw the histogram on these axes
ax.plot(hist, label="Random Hist", labelfmt="F")
# Draw the fit function
sqroot.SetNpx(1000)
ax.plot(sqroot, label="Fit", labelfmt="L", linecolor=root.kRed + 1)
# Add extra space at top of plot to make room for labels
ax.add_margins(top=0.16)
# Set axis titles
ax.set_xlabel("X [GeV]")
ax.set_ylabel("Events / 0.2 GeV")
# Add the ATLAS Label
aplt.atlas_label(text="Internal", loc="upper left")
ax.text(0.2, 0.86, "#sqrt{s} = 13 TeV, 139 fb^{-1}", size=22, align=13)
# Add legend
ax.legend(loc=(0.65, 0.8, 0.95, 0.92))
# Save the plot as a PDF
fig.savefig("hist.pdf")
def main():
# Set the ATLAS Style
aplt.set_atlas_style()
# Create a figure and axes
fig, (ax1, ax2) = aplt.ratio_plot(name="fig1",
figsize=(800, 800),
hspace=0.05)
# Define a distribution
sqroot = root.TF1("sqroot", "x*gaus(0) + [3]*abs(sin(x)/x)", 0, 10)
sqroot.SetParameters(10, 4, 1, 20)
# Randomly fill the histogram according to the above distribution
hist = root.TH1F("hist", "Random Histogram", 50, 0, 10)
hist.FillRandom("sqroot", 20000)
# Fit the histogram with the original distribution; store graphics func but do not draw
hist.Fit("sqroot", "0")
# Draw the histogram on these axes
ax1.plot(hist, "EP X0", linewidth=1, label="Random Hist", labelfmt="EP")
# Draw the fit function
sqroot.SetNpx(1000)
ax1.plot(sqroot, linecolor=root.kRed + 1, label="Fit", labelfmt="L")
# Draw line at y=0 in residuals panel
line = root.TLine(ax1.get_xlim()[0], 0, ax1.get_xlim()[1], 0)
ax2.plot(line)
# Calculate and draw the fit residuals
resids = hist.Clone("resids")
for i in range(1, resids.GetNbinsX()):
resids.SetBinContent(
i,
hist.GetBinContent(i) - sqroot.Eval(hist.GetBinCenter(i)))
resids.SetBinError(i, hist.GetBinError(i))
ax2.plot(resids, "EP X0", linewidth=1)
# Add extra space at top of plot to make room for labels
ax1.add_margins(top=0.16)
# Add extra space at top and bottom of residuals panel
ax2.add_margins(top=0.1, bottom=0.1)
# Set axis titles
ax2.set_xlabel("X [GeV]")
ax1.set_ylabel("Events / 0.2 GeV")
ax2.set_ylabel("Data - Fit", loc="center")
# Go back to top axes to add labels
ax1.cd()
# Add the ATLAS Label
aplt.atlas_label(text="Internal", loc="upper left")
ax1.text(0.2, 0.84, "#sqrt{s} = 13 TeV, 139 fb^{-1}", size=22, align=13)
# Add legend
ax1.legend(loc=(0.65, 0.78, 0.92, 0.90))
# Save the plot as a PDF
fig.savefig("fit_and_resids.pdf")
def build_reader():
reader = root.TMVA.Reader("reader")
var_mJJ = array('f',[0])
var_deltaYJJ = array('f',[0])
var_metPt = array('f',[0])
var_ptBalance = array('f',[0])
var_subleadJetEta = array('f',[0])
var_leadJetPt = array('f',[0])
var_photonEta = array('f',[0])
var_ptBalanceRed = array('f',[0])
var_nJets = array('f',[0])
var_sinDeltaPhiJJOver2 = array('f',[0])
var_deltaYJPh = array('f',[0])
reader.AddVariable("mJJ", var_mJJ)
reader.AddVariable("deltaYJJ", var_deltaYJJ)
reader.AddVariable("metPt", var_metPt)
reader.AddVariable("ptBalance", var_ptBalance)
reader.AddVariable("subleadJetEta", var_subleadJetEta)
reader.AddVariable("leadJetPt", var_leadJetPt)
reader.AddVariable("photonEta", var_photonEta)
reader.AddVariable("ptBalanceRed", var_ptBalanceRed)
reader.AddVariable("nJets", var_nJets)
reader.AddVariable("sinDeltaPhiJJOver2",var_sinDeltaPhiJJOver2)
reader.AddVariable("deltaYJPh", var_deltaYJPh)
reader.BookMVA("MLP", "dataloader/weights/TMVAClassification_MLP.weights.xml")
SDataframe, BDataframe = extract()
SDataframe["BDToutput"] = 0.0
BDataframe["BDToutput"] = 0.0
output = []
for i, row in SDataframe.iterrows():
var_mJJ[0] = row["mJJ"]
var_deltaYJJ[0] = row["deltaYJJ"]
var_metPt[0] = row["metPt"]
var_ptBalance[0] = row["ptBalance"]
var_subleadJetEta[0] = row["subleadJetEta"]
var_leadJetPt[0] = row["leadJetPt"]
var_photonEta[0] = row["photonEta"]
var_ptBalanceRed[0] = row["ptBalanceRed"]
var_nJets[0] = row["nJets"]
var_sinDeltaPhiJJOver2[0] = row["sinDeltaPhiJJOver2"]
var_deltaYJPh[0] = row["deltaYJPh"]
output.append(reader.EvaluateMVA("MLP"))
SDataframe["BDToutput"] = output
output = []
for i, row in BDataframe.iterrows():
var_mJJ[0] = row["mJJ"]
var_deltaYJJ[0] = row["deltaYJJ"]
var_metPt[0] = row["metPt"]
var_ptBalance[0] = row["ptBalance"]
var_subleadJetEta[0] = row["subleadJetEta"]
var_leadJetPt[0] = row["leadJetPt"]
var_photonEta[0] = row["photonEta"]
var_ptBalanceRed[0] = row["ptBalanceRed"]
var_nJets[0] = row["nJets"]
var_sinDeltaPhiJJOver2[0] = row["sinDeltaPhiJJOver2"]
var_deltaYJPh[0] = row["deltaYJPh"]
output.append(reader.EvaluateMVA("MLP"))
BDataframe["BDToutput"] = output
print("После отборов классификатора")
print(sum(SDataframe[SDataframe["BDToutput"] > 0.682]["weightModified"]))
print(sum(BDataframe[BDataframe["BDToutput"] > 0.682]["weightModified"]))
print("Фиксированные")
df = SDataframe[SDataframe["phCentrality"] < 0.455]
df = df[df["mJJ"] > 697]
df = df[df["ptBalance"] < 0.064]
df = df[df["deltaYJJ"] > 2.227]
print(sum(df["weightModified"]))
df = BDataframe[BDataframe["phCentrality"] < 0.455]
df = df[df["mJJ"] > 697]
df = df[df["ptBalance"] < 0.064]
df = df[df["deltaYJJ"] > 2.227]
print(sum(df["weightModified"]))
# import matplotlib.pyplot as plt
# fig, ax = plt.subplots()
# ax.hist(SDataframe["BDToutput"], weights=SDataframe["weightModified"], bins=50, color="blue", alpha=0.5, label="signal")
# ax.hist(BDataframe["BDToutput"], weights=BDataframe["weightModified"], bins=50, color="red", alpha=0.5, label="background")
# ax.legend()
# ax.set_xlabel("BDToutput")
# ax.set_ylabel("N of events")
# plt.show()
# canvas = root.TCanvas("canvas", "CANVAS", 1920, 1080)
SHist = root.TH1F("", "", 50, 0, 1)
BHist = root.TH1F("", "", 50, 0, 1)
BHist.SetStats(False)
SHist.SetStats(False)
BHist.SetLineWidth(2)
BHist.SetLineColor(2)
BHist.SetFillColor(2)
BHist.SetFillStyle(3004)
BHist.GetXaxis().CenterTitle()
BHist.GetYaxis().SetTitle("Fraction of events")
BHist.GetYaxis().CenterTitle()
BHist.GetXaxis().SetTitleOffset(1.2)
BHist.SetMinimum(0)
SHist.SetLineWidth(2)
SHist.SetLineColor(4)
SHist.SetFillColorAlpha(4, 0.2)
SHist.GetXaxis().CenterTitle()
SHist.GetYaxis().SetTitle("Fraction of events")
SHist.GetYaxis().CenterTitle()
SHist.GetXaxis().SetTitleOffset(1.2)
SHist.SetMinimum(0)
for out, weight in zip(SDataframe["BDToutput"], SDataframe["weightModified"]):
SHist.Fill(out, weight)
for out, weight in zip(BDataframe["BDToutput"], BDataframe["weightModified"]):
BHist.Fill(out, weight)
aplt.set_atlas_style()
fig, ax = aplt.subplots(1, 1, name="fig1", figsize=(800, 600))
#####################################
ax.plot(SHist, "E1")
ax.plot(BHist, "E1")
######################################
ax.add_margins(top=0.16)
ax.set_xlabel("MLP classifier response")
ax.set_ylabel("Fraction of events")
# aplt.atlas_label(text="Internal", loc="upper left")
ax.text(0.2, 0.92, "#sqrt{s} = 13 TeV, 139 fb^{-1}", size=27, align=13)
# ax.text(0.205, 0.865, "Z_{#gamma} inc. region", size=27, align=13)
legend = root.TLegend(0.65, 0.8, 0.95, 0.92)
legend.SetFillColorAlpha(0, 0)
legend.AddEntry(SHist, "Signal", "F")
legend.AddEntry(BHist, "Background", "F")
legend.Draw()
# fig.savefig("histogramms/{}.pdf".format("BDToutput"))
# input()
return make_selections(SDataframe, BDataframe)
def build_reader(model_id):
reader = root.TMVA.Reader("reader")
var_mJJ = array('f', [0])
var_deltaYJJ = array('f', [0])
var_metPt = array('f', [0])
var_ptBalance = array('f', [0])
var_subleadJetEta = array('f', [0])
var_leadJetPt = array('f', [0])
var_photonEta = array('f', [0])
var_ptBalanceRed = array('f', [0])
var_nJets = array('f', [0])
var_sinDeltaPhiJJOver2 = array('f', [0])
var_deltaYJPh = array('f', [0])
reader.AddVariable("mJJ", var_mJJ)
reader.AddVariable("deltaYJJ", var_deltaYJJ)
reader.AddVariable("metPt", var_metPt)
reader.AddVariable("ptBalance", var_ptBalance)
reader.AddVariable("subleadJetEta", var_subleadJetEta)
reader.AddVariable("leadJetPt", var_leadJetPt)
reader.AddVariable("photonEta", var_photonEta)
reader.AddVariable("ptBalanceRed", var_ptBalanceRed)
reader.AddVariable("nJets", var_nJets)
reader.AddVariable("sinDeltaPhiJJOver2", var_sinDeltaPhiJJOver2)
reader.AddVariable("deltaYJPh", var_deltaYJPh)
reader.BookMVA(
"MLP",
f"dataloader/weights/TMVAClassification_MLP{model_id}.weights.xml")
SDataframe, BDataframe = extract()
SDataframe["BDToutput"] = 0.0
BDataframe["BDToutput"] = 0.0
print("Веса до применения классификатора:")
print(np.sum(SDataframe["weightModified"]))
print(np.sum(BDataframe["weightModified"]))
output = []
for i, row in SDataframe.iterrows():
var_mJJ[0] = row["mJJ"]
var_deltaYJJ[0] = row["deltaYJJ"]
var_metPt[0] = row["metPt"]
var_ptBalance[0] = row["ptBalance"]
var_subleadJetEta[0] = row["subleadJetEta"]
var_leadJetPt[0] = row["leadJetPt"]
var_photonEta[0] = row["photonEta"]
var_ptBalanceRed[0] = row["ptBalanceRed"]
var_nJets[0] = row["nJets"]
var_sinDeltaPhiJJOver2[0] = row["sinDeltaPhiJJOver2"]
var_deltaYJPh[0] = row["deltaYJPh"]
output.append(reader.EvaluateMVA("MLP"))
SDataframe["BDToutput"] = output
output = []
for i, row in BDataframe.iterrows():
var_mJJ[0] = row["mJJ"]
var_deltaYJJ[0] = row["deltaYJJ"]
var_metPt[0] = row["metPt"]
var_ptBalance[0] = row["ptBalance"]
var_subleadJetEta[0] = row["subleadJetEta"]
var_leadJetPt[0] = row["leadJetPt"]
var_photonEta[0] = row["photonEta"]
var_ptBalanceRed[0] = row["ptBalanceRed"]
var_nJets[0] = row["nJets"]
var_sinDeltaPhiJJOver2[0] = row["sinDeltaPhiJJOver2"]
var_deltaYJPh[0] = row["deltaYJPh"]
output.append(reader.EvaluateMVA("MLP"))
BDataframe["BDToutput"] = output
SHist = root.TH1F("", "", 50, 0, 1)
BHist = root.TH1F("", "", 50, 0, 1)
BHist.SetStats(False)
SHist.SetStats(False)
BHist.SetLineWidth(2)
BHist.SetLineColor(2)
BHist.SetFillColor(2)
BHist.SetFillStyle(3004)
BHist.GetXaxis().CenterTitle()
BHist.GetYaxis().SetTitle("Fraction of events")
BHist.GetYaxis().CenterTitle()
BHist.GetXaxis().SetTitleOffset(1.2)
BHist.SetMinimum(0)
SHist.SetLineWidth(2)
SHist.SetLineColor(4)
SHist.SetFillColorAlpha(4, 0.2)
SHist.GetXaxis().CenterTitle()
SHist.GetYaxis().SetTitle("Fraction of events")
SHist.GetYaxis().CenterTitle()
SHist.GetXaxis().SetTitleOffset(1.2)
SHist.SetMinimum(0)
for out, weight in zip(SDataframe["BDToutput"],
SDataframe["weightModified"]):
SHist.Fill(out, weight)
for out, weight in zip(BDataframe["BDToutput"],
BDataframe["weightModified"]):
BHist.Fill(out, weight)
aplt.set_atlas_style()
fig, ax = aplt.subplots(1, 1, name="fig1", figsize=(800, 600))
#####################################
ax.plot(SHist, "E1")
ax.plot(BHist, "E1")
######################################
ax.add_margins(top=0.16)
# ax.add_margins(top=0.2)
ax.set_xlabel(f"MLP{model_id} classifier response")
ax.set_ylabel("Fraction of events")
ax.text(0.2, 0.92, "#sqrt{s} = 13 TeV, 139 fb^{-1}", size=27, align=13)
legend = root.TLegend(0.65, 0.8, 0.95, 0.92)
legend.SetFillColorAlpha(0, 0)
legend.AddEntry(SHist, "Signal", "F")
legend.AddEntry(BHist, "Background", "F")
legend.Draw()
fig.savefig(f"MLP{model_id}output.pdf")
# input()
return make_selections(SDataframe, BDataframe)
def significance_plot(SDataframe,
BDataframe,
ratio,
ndots=1000,
model_id="",
output_name="output",
ROC=True):
root.TMVA.Tools.Instance()
aplt.set_atlas_style()
Min = np.max(
(np.min(SDataframe[output_name]), np.min(BDataframe[output_name])))
Max = np.min(
(np.max(SDataframe[output_name]), np.max(BDataframe[output_name])))
SWSum = np.sum(SDataframe["weightModified"])
BWSum = np.sum(BDataframe["weightModified"])
XData = np.linspace(Min, Max, ndots)
SData, BData = np.array([]), np.array([])
for cursor in XData:
S = np.sum(
SDataframe[SDataframe[output_name] >= cursor]["weightModified"])
B = np.sum(
BDataframe[BDataframe[output_name] >= cursor]["weightModified"])
SData = np.append(SData, S)
BData = np.append(BData, B)
YData = SData / np.sqrt(ratio * (SData + BData))
SEff = SData / SWSum
BRej = 1 - BData / BWSum
XPlot, YPlot = array("d"), array("d")
for x, y in zip(XData, YData):
XPlot.append(x)
YPlot.append(y)
curve = root.TGraph(ndots, XPlot, YPlot)
curve.SetLineColor(2)
curve.SetLineWidth(2)
curve.SetMarkerColor(2)
curve.SetMarkerSize(0)
curve.GetXaxis().SetRangeUser(Min, Max)
curve.GetXaxis().SetTitle(f"Cut value applied on BDTgrad{model_id} output")
curve.GetYaxis().SetTitle('Significance')
fig, ax = aplt.subplots(1, 1, name="", figsize=(800, 600))
ax.plot(curve)
ax.add_margins(top=0.16)
peak_index = YData.argmax()
cut = XData[peak_index]
sig_max = YData[peak_index]
ax.text(0.2, 0.92, "#sqrt{s} = 13 TeV, 139 fb^{-1}", size=27, align=13)
text1 = "For {} signal and {} background".\
format(round(np.sum(SDataframe["weightModified"])),
round(np.sum(BDataframe["weightModified"])))
text2 = "events the maximum {} is".format("S/#sqrt{S+B}")
text3 = "{} when cutting at {}".format(round(sig_max, 3), round(cut, 3))
ax.text(0.2, 0.3, text1, size=20, align=13)
ax.text(0.2, 0.27, text2, size=20, align=13)
ax.text(0.2, 0.22, text3, size=20, align=13)
line = root.TLine(cut, 0, cut, 4)
line.SetLineStyle(10)
line.SetLineColor(6)
ax.plot(line)
fig.savefig(f"BDTgrad_{model_id}_outputCut.pdf")
SWCut = np.array(
SDataframe[SDataframe[output_name] > cut]["weightModified"])
BWCut = np.array(
BDataframe[BDataframe[output_name] > cut]["weightModified"])
print("Significance error:")
print(error(SWCut, BWCut) /
np.sqrt(ratio)) ### Нужно ли делить на sqrt(ratio)?
if ROC:
ROC_plot(SEff, BRej, model_id=model_id)
def setup(fontsize=27): root.TMVA.Tools.Instance() aplt.set_atlas_style(fontsize) root.gStyle.SetErrorX(0.5) root.gStyle.SetEndErrorSize(0.1)
import ROOT as root
import numpy as np
import atlasplots as aplt
aplt.set_atlas_style()
fig, ax = aplt.subplots(1, 1, name="fig", figsize=(800, 800))
print(dir(ax.get_ylabel()))
input()
# c1 = root.TCanvas("c1","hists with different scales",600,400)
hist = root.TH1F("", "", 20, -1, 1)
a = np.random.randn(1000, 1).ravel()
for el in a:
hist.Fill(el, 0.5)
width = hist.GetBinWidth(1)
root.gStyle.SetErrorX(0.5)
ax.plot(hist, [-1, 1, 0, 30], options="E1")
# hist.Draw("E1")
input()
def main():
# Set the ATLAS Style
aplt.set_atlas_style()
# Create a figure and axes
fig, (ax1, ax2) = aplt.ratio_plot(name="fig1",
figsize=(800, 800),
hspace=0.05)
# Define "background", "signal" and "data" distributions
bkg_func = root.TF1("bkg_func", "[0]*exp([1]*x)", 0, 10)
bkg_func.SetParameters(20, -0.5)
sig_func = root.TF1("sig_func", "x*gaus(0)", 0, 10)
sig_func.SetParameters(10, 4, 1)
# Randomly fill histograms according to the above distributions
bkg_hist = root.TH1F("bkg_hist", "Background", 50, 0, 10)
bkg_hist.FillRandom("bkg_func", 10000)
bkg_hist.SetFillColor(root.kRed + 1)
bkg_hist.SetLineWidth(0)
bkg_hist.Sumw2()
sig_hist = root.TH1F("sig_hist", "Signal", 50, 0, 10)
sig_hist.FillRandom("sig_func", 10000)
sig_hist.SetFillColor(root.kAzure + 1)
sig_hist.SetLineWidth(0)
sig_hist.Sumw2()
data_hist = root.TH1F("data_hist", "Data", 50, 0, 10)
data_hist.FillRandom("bkg_func", 10000)
data_hist.FillRandom("sig_func", 10000)
data_hist.Sumw2()
data_hist.SetBinErrorOption(
root.TH1.EBinErrorOpt.kPoisson) # Use 68% Poisson errors
# Stack the background and signal histograms
bkg_and_sig = root.THStack("bkg_and_sig", "")
bkg_and_sig.Add(bkg_hist)
bkg_and_sig.Add(sig_hist)
# Draw the stacked histogram on these axes
ax1.plot(bkg_and_sig)
# Plot the MC stat error as a hatched band
err_band = aplt.root_helpers.hist_to_graph(bkg_and_sig.GetStack().Last(),
show_bin_width=True)
ax1.plot(err_band, "2", fillcolor=1, fillstyle=3254, linewidth=0)
# ax1.set_yscale("log") # uncomment to use log scale for y axis
# Plot the data as a graph
data_graph = aplt.root_helpers.hist_to_graph(data_hist)
ax1.plot(data_graph, "P")
# Use same x-range in lower axes as upper axes
ax2.set_xlim(ax1.get_xlim())
# Draw line at y=1 in ratio panel
line = root.TLine(ax1.get_xlim()[0], 1, ax1.get_xlim()[1], 1)
ax2.plot(line)
# Plot the relative error on the ratio axes
err_band_ratio = aplt.root_helpers.hist_to_graph(
bkg_and_sig.GetStack().Last(), show_bin_width=True, norm=True)
ax2.plot(err_band_ratio, "2", fillcolor=1, fillstyle=3254)
# Calculate and draw the ratio
ratio_hist = data_hist.Clone("ratio_hist")
ratio_hist.Divide(bkg_and_sig.GetStack().Last())
ratio_graph = aplt.root_helpers.hist_to_graph(ratio_hist)
ax2.plot(ratio_graph, "P0")
# Add extra space at top of plot to make room for labels
ax1.add_margins(top=0.16)
# Set axis titles
ax2.set_xlabel("X [GeV]")
ax1.set_ylabel("Events / 0.2 GeV")
ax2.set_ylabel("Data / Pred.", loc="centre")
ax2.set_ylim(0.75, 1.25)
ax2.draw_arrows_outside_range(ratio_graph)
# Go back to top axes to add labels
ax1.cd()
# Add the ATLAS Label
aplt.atlas_label(text="Internal", loc="upper left")
ax1.text(0.2, 0.84, "#sqrt{s} = 13 TeV, 139 fb^{-1}", size=22, align=13)
# Add legend
legend = ax1.legend(loc=(0.68, 0.65, 1 - root.gPad.GetRightMargin() - 0.03,
1 - root.gPad.GetTopMargin() - 0.03),
textsize=22)
legend.AddEntry(data_graph, "Data", "EP")
legend.AddEntry(sig_hist, "Signal", "F")
legend.AddEntry(bkg_hist, "Background", "F")
legend.AddEntry(err_band, "MC Stat. Unc.", "F")
# Save the plot as a PDF
fig.savefig("data_vs_mc.pdf")