-
Notifications
You must be signed in to change notification settings - Fork 0
/
XGB_apply_psiOpt.py
840 lines (810 loc) · 41.6 KB
/
XGB_apply_psiOpt.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
# -*- coding: utf-8 -*-
import argparse
import root_numpy
import ROOT
import numpy as np
import pandas as pd
import os.path
import matplotlib
matplotlib.use('Agg')
import matplotlib.pyplot as plt
import yaml
import xgboost as xgb
from math import sqrt
from sklearn.cross_validation import StratifiedKFold
ROOT.gROOT.SetBatch(True)
def fit_doubleCB(a_mc_x, a_data, out_path, s_info='', save_fig=True):
# Initialise dictionary for storing fit info
d_fit_info = {}
# Format arrays
a_fit_mc = a_mc_x.astype(dtype=[('b_mass_mc', np.float)])
a_fit_data = a_data.astype(dtype=[('b_mass' , np.float)])
# Estimate sig/bkg yields
max_yield_est = a_fit_data.shape[0] # max possible signal yield
bkg_yield_est = len(a_fit_data[a_fit_data > 5350]) * 4. # estimate background from sideband
# Create tree for fitting
t_fit_data = root_numpy.array2tree(a_fit_data)
t_fit_mc = root_numpy.array2tree(a_fit_mc)
## Monte Carlo
# Inialise parameters for fit
b_mass_mc = ROOT.RooRealVar("b_mass_mc" , "B mass MC [MeV]", 5220. , 5400.)
mean_mc = ROOT.RooRealVar("mean_mc" , "mean_mc" , 5279. , 5195. , 5400.)
sig_1_mc = ROOT.RooRealVar("sig_1_mc" , "sig_1_mc" , 5. , .1 , 15.)
alpha_1_mc = ROOT.RooRealVar("alpha_1_mc" , "alpha_1_mc" , 5. , .1 , 10.)
n_1_mc = ROOT.RooRealVar("n_1_mc" , "n_1_mc" , 2. , 0. , 15.)
r_s1_s2_mc = ROOT.RooRealVar("r_s1_s2_mc" , "r_s1_s2_mc" , .5, .1 , 10.)
sig_2_mc = ROOT.RooFormulaVar("sig_2_mc" , "sig_1_mc*r_s1_s2_mc", ROOT.RooArgList(sig_1_mc, r_s1_s2_mc))
alpha_2_mc = ROOT.RooRealVar("alpha_2_mc" , "alpha_2_mc" , -5. , -10. , -.1)
n_2_mc = ROOT.RooRealVar("n_2_mc" , "n_2_mc" , 2. , 0. , 15.)
r_cb1_cb2_mc = ROOT.RooRealVar("r_cb1_cb2_mc", "r_cb1_cb2_mc" , .5, .01, 1.)
# Initialise fit model
cb_1_mc = ROOT.RooCBShape("cb_1_mc" , "cb_1_mc" , b_mass_mc, mean_mc, sig_1_mc, alpha_1_mc, n_1_mc)
cb_2_mc = ROOT.RooCBShape("cb_2_mc" , "cb_2_mc" , b_mass_mc, mean_mc, sig_2_mc, alpha_2_mc, n_2_mc)
model_sig_mc = ROOT. RooAddPdf("model_sig_mc", "model_sig_mc", ROOT.RooArgList(cb_1_mc, cb_2_mc), ROOT.RooArgList(r_cb1_cb2_mc))
# Initialise dataset
dataset_mc = ROOT.RooDataSet("dataset_mc","dataset from tree", t_fit_mc, ROOT.RooArgSet(b_mass_mc))
# Perform fit
mean_mc .setConstant(ROOT.kFALSE)
sig_1_mc .setConstant(ROOT.kFALSE)
alpha_1_mc .setConstant(ROOT.kFALSE)
n_1_mc .setConstant(ROOT.kFALSE)
r_s1_s2_mc .setConstant(ROOT.kFALSE)
alpha_2_mc .setConstant(ROOT.kFALSE)
n_2_mc .setConstant(ROOT.kFALSE)
r_cb1_cb2_mc.setConstant(ROOT.kFALSE)
model_sig_mc.fitTo(dataset_mc, ROOT.RooFit.Range(5220., 5400.))
# Store fitted values
f_mean_mc = ROOT.RooRealVar("f_mean_mc" , "f_mean_mc" , mean_mc .getValV())
f_sig_1_mc = ROOT.RooRealVar("f_sig_1_mc" , "f_sig_1_mc" , sig_1_mc .getValV())
f_alpha_1_mc = ROOT.RooRealVar("f_alpha_1_mc", "f_alpha_1_mc", alpha_1_mc.getValV())
f_n_1_mc = ROOT.RooRealVar("f_n_1_mc" , "f_n_1_mc" , n_1_mc .getValV())
f_sig_2_mc = ROOT.RooRealVar("f_sig_2_mc" , "f_sig_2_mc" , sig_2_mc .getValV())
f_alpha_2_mc = ROOT.RooRealVar("f_alpha_2_mc", "f_alpha_2_mc", alpha_2_mc.getValV())
f_n_2_mc = ROOT.RooRealVar("f_n_2_mc" , "f_n_2_mc" , n_1_mc .getValV())
# Store fitted models
f_cb_1_mc = ROOT.RooCBShape("cb_1_mc" , "cb_1_mc" , b_mass_mc, f_mean_mc, f_sig_1_mc, f_alpha_1_mc, f_n_1_mc)
f_cb_2_mc = ROOT.RooCBShape("cb_2_mc" , "cb_2_mc" , b_mass_mc, f_mean_mc, f_sig_2_mc, f_alpha_2_mc, f_n_2_mc)
f_model_sig_mc = ROOT. RooAddPdf("model_sig_mc", "model_sig_mc", ROOT.RooArgList(f_cb_1_mc, f_cb_2_mc), ROOT.RooArgList(r_cb1_cb2_mc))
## Plot
# Frame for fit
frame_mc = b_mass_mc.frame()
# Frame for pulls
frame_mc_pull = b_mass_mc.frame()
# Add data and fit to frame
dataset_mc.plotOn(frame_mc)
f_model_sig_mc.plotOn(frame_mc)
# Plot on split canvas
c = ROOT.TCanvas("X3872MC", "X3872MC", 400, 500)
c.Divide(1, 2, 0, 0)
# Plot data and fit
c.cd(2)
ROOT.gPad.SetTopMargin(0)
ROOT.gPad.SetLeftMargin(0.15)
ROOT.gPad.SetRightMargin(0.035)
ROOT.gPad.SetPad(.01,.01,.95,.77)
frame_mc.SetTitle("Fitted Monte-Carlo Bmass")
frame_mc.SetMaximum(frame_mc.GetMaximum()*1.1)
frame_mc.GetYaxis().SetTitleOffset(1.6)
frame_mc.Draw()
# Plot pulls
c.cd(1)
ROOT.gPad.SetTopMargin(0)
ROOT.gPad.SetLeftMargin(0.15)
ROOT.gPad.SetRightMargin(0.035)
ROOT.gPad.SetPad(.01,.76,.95,.97)
# Determine pulls and format
h_pull_mc = frame_mc.pullHist()
h_pull_mc.SetFillColor(15)
h_pull_mc.SetFillStyle(3144)
# Add pulls to frame
frame_mc_pull.addPlotable(h_pull_mc,'L3')
frame_mc_pull.GetYaxis().SetNdivisions(505)
frame_mc_pull.GetYaxis().SetLabelSize(0.20)
frame_mc_pull.SetTitle("")
frame_mc_pull.Draw()
# Save canvas
if save_fig:
c.SaveAs(out_path+s_info+'_FittedMassDistribution_MonteCarlo.pdf')
# Store fit variables
d_fit_info['mc_mean'] = mean_mc.getValV()
d_fit_info['mc_mean_err'] = mean_mc.getError()
d_fit_info['mc_sig1'] = sig_1_mc.getValV()
d_fit_info['mc_sig1_err'] = sig_1_mc.getError()
d_fit_info['mc_r_s1_s2'] = r_s1_s2_mc.getValV()
d_fit_info['mc_r_s1_s2_err'] = r_s1_s2_mc.getError()
d_fit_info['mc_alpha1'] = alpha_1_mc.getValV()
d_fit_info['mc_alpha1_err'] = alpha_1_mc.getError()
d_fit_info['mc_alpha2'] = alpha_2_mc.getValV()
d_fit_info['mc_alpha2_err'] = alpha_2_mc.getError()
d_fit_info['mc_n1'] = n_1_mc.getValV()
d_fit_info['mc_n1_err'] = n_1_mc.getError()
d_fit_info['mc_n2'] = n_2_mc.getValV()
d_fit_info['mc_n2_err'] = n_2_mc.getError()
d_fit_info['mc_r_cb1_cb2'] = r_cb1_cb2_mc.getValV()
d_fit_info['mc_r_cb1_cb2_err'] = r_cb1_cb2_mc.getError()
d_fit_info['mc_fit_chi2'] = frame_mc.chiSquare()
## Data
b_mass = ROOT.RooRealVar("b_mass" , "B mass [MeV]" , 5220., 5400.)
mean = ROOT.RooRealVar("mean" , "mean" , mean_mc .getValV(), 5220., 5400.)
sig_1 = ROOT.RooRealVar("sig_1" , "sig_1" , sig_1_mc .getValV(), .1, 15.)
alpha_1 = ROOT.RooRealVar("alpha_1" , "alpha_1" , alpha_1_mc .getValV(), .1, 10.)
n_1 = ROOT.RooRealVar("n_1" , "n_1" , n_1_mc .getValV(), 0., 15.)
r_s1_s2 = ROOT.RooRealVar("r_s1_s2" , "r_s1_s2" , r_s1_s2_mc .getValV(), 0.01, 10)
sig_2 = ROOT.RooFormulaVar("sig_2" , "sig_1*r_s1_s2", ROOT.RooArgList(sig_1, r_s1_s2))
alpha_2 = ROOT.RooRealVar("alpha_2" , "alpha_2" , alpha_2_mc .getValV(), -10., -.1)
n_2 = ROOT.RooRealVar("n_2" , "n_2" , n_2_mc .getValV(), 0., 15.)
r_cb1_cb2 = ROOT.RooRealVar("r_cb1_cb2", "r_cb1_cb2" , r_cb1_cb2_mc.getValV(), 0.01 , 1.);
sig_yield = ROOT.RooRealVar("sig_yield", "sig_yield" , 0, (max_yield_est - bkg_yield_est)*1.5)
# Background
exp_c = ROOT.RooRealVar("exp_c" , "exp_c" , 0., -.02, .02)
bgr_yield = ROOT.RooRealVar("bgr_yield" , "bgr_yield", bkg_yield_est*.5, max_yield_est)
# Initialise fit model
cb_1 = ROOT.RooCBShape("cb_1" , "cb_1" , b_mass, mean, sig_1, alpha_1, n_1)
cb_2 = ROOT.RooCBShape("cb_2" , "cb_2" , b_mass, mean, sig_2, alpha_2, n_2)
exp_bg = ROOT.RooExponential("exp_bg", "exp_bg", b_mass, exp_c)
model_sig = ROOT.RooAddPdf("model_sig", "model_sig", ROOT.RooArgList(cb_1, cb_2) , ROOT.RooArgList(r_cb1_cb2))
model_tot = ROOT.RooAddPdf("model_tot", "model_tot", ROOT.RooArgList(model_sig, exp_bg), ROOT.RooArgList(sig_yield, bgr_yield))
# Initialise dataset
dataset = ROOT.RooDataSet("dataset","dataset from tree", t_fit_data, ROOT.RooArgSet(b_mass))
# Perform fit - kTRUE vars determined from MC fit
mean .setConstant(ROOT.kFALSE)
sig_1 .setConstant(ROOT.kFALSE)
alpha_1 .setConstant(ROOT.kTRUE)
n_1 .setConstant(ROOT.kTRUE)
r_s1_s2 .setConstant(ROOT.kTRUE)
alpha_2 .setConstant(ROOT.kTRUE)
n_2 .setConstant(ROOT.kTRUE)
r_cb1_cb2.setConstant(ROOT.kTRUE)
sig_yield.setConstant(ROOT.kFALSE)
exp_c .setConstant(ROOT.kFALSE)
bgr_yield.setConstant(ROOT.kFALSE)
model_tot.fitTo(dataset, ROOT.RooFit.Range(5220., 5380.))
# Store fitted values
f_mean = ROOT.RooRealVar("f_mean" , "f_mean" , mean .getValV())
f_sig_1 = ROOT.RooRealVar("f_sig_1" , "f_sig_1" , sig_1 .getValV())
f_alpha_1 = ROOT.RooRealVar("f_alpha_1" , "f_alpha_1" , alpha_1 .getValV())
f_n_1 = ROOT.RooRealVar("f_n_1" , "f_n_1" , n_1 .getValV())
f_sig_2 = ROOT.RooRealVar("f_sig_2" , "f_sig_2" , sig_2 .getValV())
f_alpha_2 = ROOT.RooRealVar("f_alpha_2" , "f_alpha_2" , alpha_2 .getValV())
f_n_2 = ROOT.RooRealVar("f_n_2" , "f_n_2" , n_1 .getValV())
f_exp_c = ROOT.RooRealVar("f_exp_c" , "f_exp_c" , exp_c .getValV())
f_sig_yield = ROOT.RooRealVar("f_sig_yield", "f_sig_yield", sig_yield.getValV())
f_bgr_yield = ROOT.RooRealVar("f_bgr_yield", "f_bgr_yield", bgr_yield.getValV())
# Store fitted models
f_cb_1 = ROOT. RooCBShape("cb_1" , "cb_1" , b_mass, f_mean, f_sig_1, f_alpha_1, f_n_1)
f_cb_2 = ROOT. RooCBShape("cb_2" , "cb_2" , b_mass, f_mean, f_sig_2, f_alpha_2, f_n_2)
f_exp_bg = ROOT.RooExponential("exp_bg" , "exp_bg" , b_mass, f_exp_c)
f_model_sig = ROOT. RooAddPdf("model_sig", "model_sig", ROOT.RooArgList(f_cb_1, f_cb_2), ROOT.RooArgList(r_cb1_cb2))
f_model_tot = ROOT. RooAddPdf("model_tot", "model_tot", ROOT.RooArgList(f_model_sig, f_exp_bg), ROOT.RooArgList(f_sig_yield, f_bgr_yield))
# Plot
# Frame for fit
frame = b_mass.frame()
# Frame for pulls
frame_pull = b_mass.frame()
# Add data and fit to frame
dataset.plotOn(frame)
f_model_tot.plotOn(frame)
# Plot on split canvas
c = ROOT.TCanvas("data_fit", "data_fit", 400, 500)
c.Divide(1, 2, 0, 0)
# Plot data and fit
c.cd(2)
ROOT.gPad.SetTopMargin(0)
ROOT.gPad.SetLeftMargin(0.15)
ROOT.gPad.SetRightMargin(0.035)
ROOT.gPad.SetPad(.01,.01,.95,.77)
frame.SetTitle("Fitted Data Bmass")
frame.SetMaximum(frame.GetMaximum()*1.1)
frame.GetYaxis().SetTitleOffset(1.6)
frame.Draw()
# Plot pulls
c.cd(1)
ROOT.gPad.SetTopMargin(0)
ROOT.gPad.SetLeftMargin(0.15)
ROOT.gPad.SetRightMargin(0.035)
ROOT.gPad.SetPad(.01,.76,.95,.97)
# Determine pulls and format
h_pull = frame.pullHist()
h_pull.SetFillColor(15)
h_pull.SetFillStyle(3144)
# Add pulls to frame
frame_pull.addPlotable(h_pull,'L3')
frame_pull.GetYaxis().SetNdivisions(505)
frame_pull.GetYaxis().SetLabelSize(0.20)
frame_pull.SetTitle("")
frame_pull.Draw()
# Save plot
if save_fig:
c.SaveAs(out_path+s_info+'_FittedMassDistribution_Data.pdf')
# Store fit variables
d_fit_info['data_mean'] = mean.getValV()
d_fit_info['data_mean_err'] = mean.getError()
d_fit_info['data_sig1'] = sig_1.getValV()
d_fit_info['data_sig1_err'] = sig_1.getError()
d_fit_info['data_r_s1_s2'] = r_s1_s2.getValV()
d_fit_info['data_r_s1_s2_err'] = r_s1_s2.getError()
d_fit_info['data_alpha1'] = alpha_1.getValV()
d_fit_info['data_alpha1_err'] = alpha_1.getError()
d_fit_info['data_alpha2'] = alpha_2.getValV()
d_fit_info['data_alpha2_err'] = alpha_2.getError()
d_fit_info['data_n1'] = n_1.getValV()
d_fit_info['data_n1_err'] = n_1.getError()
d_fit_info['data_n2'] = n_2.getValV()
d_fit_info['data_n2_err'] = n_2.getError()
d_fit_info['data_r_cb1_cb2'] = r_cb1_cb2.getValV()
d_fit_info['data_r_cb1_cb2_err'] = r_cb1_cb2.getError()
d_fit_info['data_expc'] = exp_c.getValV()
d_fit_info['data_expc_err'] = exp_c.getError()
d_fit_info['data_sig_yield'] = sig_yield.getValV()
d_fit_info['data_sig_yield_err'] = sig_yield.getError()
d_fit_info['data_bgr_yield'] = bgr_yield.getValV()
d_fit_info['data_bgr_yield_err'] = bgr_yield.getError()
d_fit_info['data_fit_chi2'] = frame.chiSquare()
# Return fit info dictionary
return d_fit_info
def main(args):
### Perpare data for processing
# Ensure output directory exists
out_path = args.data_dir+'results/'
if not os.path.exists(out_path):
os.makedirs(out_path)
# Open config file
f_config = open(args.config)
global D_CONFIGS
D_CONFIGS = yaml.load(f_config)
## Load files
# List required branches (only once)
s_load_branches = set()
for run in list(D_CONFIGS.keys()):
s_load_branches |= set(D_CONFIGS[run]['fit_vars'])
s_load_branches.add('prob_'+run)
s_load_branches.add('scaledmass')
s_load_branches.add('mjpipi')
s_load_branches.add('bplus_PT')
l_load_branches = list(s_load_branches)
# Load files into arrays
print('*** Loading Data ***')
a_mc_x = root_numpy.root2array(args.data_dir+'mc_x_proba.root', treename = args.tree_name, branches = l_load_branches)
a_mc_p = root_numpy.root2array(args.data_dir+'mc_p_proba.root', treename = args.tree_name, branches = l_load_branches)
a_side = root_numpy.root2array(args.data_dir+'side_proba.root', treename = args.tree_name, branches = l_load_branches)
a_data = root_numpy.root2array(args.data_dir+'data_proba.root', treename = args.tree_name, branches = l_load_branches)
print('*** Processing Data ***')
# Convert to DataFrames
df_mc_x = pd.DataFrame(a_mc_x)
df_mc_p = pd.DataFrame(a_mc_p)
df_side = pd.DataFrame(a_side)
df_data = pd.DataFrame(a_data)
# Add categoriastion
df_mc_x['cat'] = 'mc_x'
df_mc_p['cat'] = 'mc_p'
df_side['cat'] = 'side'
# Add target
df_mc_x['class'] = 1
df_mc_p['class'] = 1
df_side['class'] = 0
# Combine into training set
df_train = pd.concat([df_mc_x, df_mc_p, df_side])
# Print summary stats
print(' *** Data loaded ***')
print(' *** Training events: %d ***'%(df_train.shape[0]))
print(' *** Data events: %d ***'%(df_data .shape[0]))
# Dictionaries for storing information on each run
d_run_info = {}
d_roc_plot = {}
### Estimate signal yield - all data
d_sig_est_alldata = fit_doubleCB(pd.concat([df_mc_x, df_mc_p])['scaledmass'].as_matrix(), df_data['scaledmass'].as_matrix(), out_path, s_info='alldata_signal_est')
d_run_info['sig_est_alldata'] = d_sig_est_alldata
for run in list(D_CONFIGS.keys()):
print('*** Performing run %s ***'%(run))
d_run_info[run] = {}
d_roc_plot[run] = {}
out_path_run = out_path+run+'/'
if not os.path.exists(out_path_run):
os.makedirs(out_path_run)
if args.opt_cut is None:
### Find optimal cut
print(' *** Determining optimal cut ***')
# Optimise the probability cut
sig_effs_mcj = [] # record signal efficiencies - on MC X(3823) only
sig_effs_mcx = [] # record signal efficiencies - on MC psi(2S) only
sig_effs_all = [] # record signal efficiencies
sig_effs_pdata = []# # record signal efficiencies on psi(2S) range data
bgr_rejs = [] # record background rejections
cut_scores = [] # record cut optimisation metric
# Determine cut metric for a range of cuts
cuts = np.linspace(.0, 1., 100, endpoint=False)
df_data_p = df_data[(df_data['mjpipi'] < 3696) & (df_data['mjpipi'] > 3676)]
s0_X = 184
s0_p = float(fit_doubleCB(df_mc_p['scaledmass'].as_matrix(), df_data_p['scaledmass'].as_matrix(), out_path, s_info='psi(2S)_s0_est')['data_sig_yield'])
for prob_threshold in cuts:
# Apply cut to psi(2S) dataframe
df_data_p_cut = df_data_p[df_data_p['prob_'+run] > prob_threshold]
df_mc_p_cut = df_mc_p[df_mc_p['prob_'+run] > prob_threshold]
d_datap_fit_cut = fit_doubleCB(df_mc_p_cut['scaledmass'].as_matrix(), df_data_p_cut['scaledmass'].as_matrix(), out_path, s_info='psi(2S)_cut_est_{:.3f}'.format(prob_threshold), save_fig=False)
# Determine how many predictions are correct
signal_efficiency_mcx = float(df_train[(df_train['prob_'+run] > prob_threshold) & (df_train['cat'] == 'mc_x')].shape[0]) / float(df_train[df_train['cat'] == 'mc_x'].shape[0])
signal_efficiency_mcj = float(df_train[(df_train['prob_'+run] > prob_threshold) & (df_train['cat'] == 'mc_p')].shape[0]) / float(df_train[df_train['cat'] == 'mc_p'].shape[0])
signal_efficiency_all = float(df_train[(df_train['prob_'+run] > prob_threshold) & (df_train['class'] == 1 )].shape[0]) / float(df_train[df_train['class'] == 1 ].shape[0])
signal_efficiency_p_data = float(d_datap_fit_cut['data_sig_yield']) / s0_p
background_rejection = float(df_train[(df_train['prob_'+run] < prob_threshold) & (df_train['class'] == 0 )].shape[0]) / float(df_train[df_train['class'] == 0 ].shape[0])
# Store scores
sig_effs_all .append(signal_efficiency_all)
sig_effs_mcj .append(signal_efficiency_mcj)
sig_effs_mcx .append(signal_efficiency_mcx)
#sig_effs_pdata.append(signal_efficiency_p_data)
bgr_rejs .append(background_rejection)
# Optimize cut
eff = signal_efficiency_p_data
# Background events, scaled to 40MeV window about B peak, considering only those in X(3823) region
B = df_train[((df_train['prob_'+run] > prob_threshold)) & ((df_train['scaledmass'] > 5400.) & (df_train['scaledmass'] < 5450.)) & ((df_train['mjpipi'] > 3773) & (df_train['mjpipi'] < 3873))].shape[0] * .8
cut_scores.append((s0_X * eff) / sqrt((s0_X * eff) + B))
# Base on cut optimisation metric
cut_index = np.argmax(cut_scores)
prob_threshold = cuts[cut_index]
### Store some parameters of interest
d_run_info[run]['optimal_cut'] = np.asscalar(prob_threshold)
#d_run_info[run]['psi_signal_efficiency'] = sig_effs_pdata[cut_index]
d_run_info[run]['all_signal_efficiency'] = sig_effs_all [cut_index]
d_run_info[run]['mcj_signal_efficiency'] = sig_effs_mcj [cut_index]
d_run_info[run]['mcx_signal_efficiency'] = sig_effs_mcx [cut_index]
d_roc_plot[run]['sig_effs'] = sig_effs_mcj
d_roc_plot[run]['bgr_rejs'] = bgr_rejs
### Print some summary stats
print(' *** Optimal cut: %1.2f ***'%(prob_threshold))
print(' *** Signal efficiency: %1.2f ***'%(d_run_info[run]['all_signal_efficiency']))
print(' *** MCJ Signal efficiency: %1.2f ***'%(d_run_info[run]['mcj_signal_efficiency']))
print(' *** MCX Signal efficiency: %1.2f ***'%(d_run_info[run]['mcx_signal_efficiency']))
else:
prob_threshold = float(args.opt_cut)
### Apply model to data
df_data['class'] = df_data['prob_'+run] > prob_threshold
### Plot cut optimisation
print(' *** Plotting cut optimisation ***')
fig = plt.figure()
plt.plot(cuts, cut_scores)
plt.ylabel("Cut Score")
plt.xlabel("Probability Threshold")
plt.xlim(0.,1.)
plt.title("Cut Score "+run)
plt.tight_layout(pad=2.0)
fig.savefig(out_path_run+'cut_score.pdf')
plt.close()
### Plot mass histogram for optimal cut
print(' *** Plotting Mass histograms ***')
# Initialise canvas
c_name = 'B_Mass_Distribution '+run
c = ROOT.TCanvas(c_name, c_name, 600, 400)
c.cd()
# Select required quantity
a_raw = df_data['scaledmass'].as_matrix()
a_cut = df_data[df_data['class']==1]['scaledmass'].as_matrix()
# Create and format histograms
h_raw = ROOT.TH1F(c_name+'_No_Cut' , c_name+'_No_Cut;B Mass [MeV/#it{c}^{2}];candidates/18[MeV/#it{c}^{2}]', 100, 5220., 5400.)
h_cut = ROOT.TH1F(c_name+'_XGB_Cut', c_name+'_XGB_Cut;B Mass [MeV/#it{c}^{2}];candidates/18[MeV/#it{c}^{2}]', 100, 5220., 5400.)
# Fill histograms
map(h_raw.Fill, a_raw)
map(h_cut.Fill, a_cut)
# Normalise
## Make it pretty
h_raw.SetTitle('B Mass Distribution '+run)
# Format for each case of x-axis
h_raw.GetYaxis().SetTitleOffset(1.6)
y_max = 1.1*max(h_raw.GetBinContent(h_raw.GetMaximumBin()), h_cut.GetBinContent(h_cut.GetMaximumBin()))
y_min = 0.9*min(h_raw.GetBinContent(h_raw.GetMinimumBin()), h_cut.GetBinContent(h_cut.GetMinimumBin()))
h_raw.GetYaxis().SetRangeUser(y_min, y_max)
# Format plotting style
h_raw.SetLineColor(ROOT.kRed)
h_raw.SetFillColorAlpha(ROOT .kRed - 10, 0.7)
h_cut.SetLineColor(ROOT.kBlue)
h_cut.SetFillColorAlpha(ROOT.kBlue - 10, 0.7)
# Remove stats boxes
h_raw.SetStats(False)
h_cut.SetStats(False)
# Print
h_raw.Draw('HIST')
h_cut.Draw('HISTsame')
# Create legend
leg = ROOT.TLegend(0.6,0.75,0.9,0.9)
leg.AddEntry(h_raw, 'Uncut Data', 'L')
leg.AddEntry(h_cut, 'XGBoost cut: {:.3f}'.format(prob_threshold), 'L')
leg.SetLineColor(0)
leg.SetLineStyle(0)
leg.SetFillStyle(0)
leg.SetBorderSize(0)
leg.Draw('same')
# Save
c.SaveAs(out_path_run+'Mass_histogram_B_XGBcut.pdf')
### BDT answer
print(' *** Plotting classification probabilities ***')
# Initialise canvas
c_name = 'BDT_Predicted_Probability '+run
c = ROOT.TCanvas(c_name, c_name, 600, 400)
c.cd()
# Select required quantity
a_train_sig_prob = df_train['prob_'+run][df_train['class'] == 1].as_matrix()
a_train_bkg_prob = df_train['prob_'+run][df_train['class'] == 0].as_matrix()
a_data_prob = df_data['prob_'+run].as_matrix()
# Create and format histograms
h_train_sig_prob = ROOT.TH1F(c_name+'_Sig_Prob' , c_name+'_Sig_Prob;Probability;Candidates', 100, 0., 1.)
h_train_bkg_prob = ROOT.TH1F(c_name+'_Bkg_Prob' , c_name+'_Bkg_Prob;Probability;Candidates', 100, 0., 1.)
h_data_prob = ROOT.TH1F(c_name+'_Data_Prob', c_name+'_Data_Prob;Probability;Candidates', 100, 0., 1.)
# Fill histograms
map(h_train_sig_prob.Fill, a_train_sig_prob)
map(h_train_bkg_prob.Fill, a_train_bkg_prob)
map(h_data_prob.Fill, a_data_prob)
# Normalise
h_train_sig_prob.Scale(1./h_train_sig_prob.Integral())
h_train_bkg_prob.Scale(1./h_train_bkg_prob.Integral())
h_data_prob .Scale(1./h_data_prob .Integral())
## Make it pretty
h_train_sig_prob.SetTitle('Event Probability Distribution '+run)
# Format for each case of x-axis
h_train_sig_prob.GetYaxis().SetTitleOffset(1.6)
y_max = 1.1*max(max(h_train_sig_prob.GetBinContent(h_train_sig_prob.GetMaximumBin()), h_train_bkg_prob.GetBinContent(h_train_bkg_prob.GetMaximumBin())), h_data_prob.GetBinContent(h_data_prob.GetMaximumBin()))
y_min = 0.9*min(min(h_train_sig_prob.GetBinContent(h_train_sig_prob.GetMinimumBin()), h_train_bkg_prob.GetBinContent(h_train_bkg_prob.GetMinimumBin())), h_data_prob.GetBinContent(h_data_prob.GetMinimumBin()))
h_train_sig_prob.GetYaxis().SetRangeUser(y_min, y_max)
# Format plotting style
h_train_sig_prob.SetLineColor(ROOT.kRed)
h_train_sig_prob.SetFillColorAlpha(ROOT .kRed - 10, 0.7)
h_train_bkg_prob.SetLineColor(ROOT.kBlue)
h_train_bkg_prob.SetFillColorAlpha(ROOT.kBlue - 10, 0.7)
h_data_prob.SetLineColor(ROOT.kGreen)
h_data_prob.SetFillColorAlpha(ROOT.kGreen - 10, 0.7)
# Remove stats boxes
h_train_sig_prob.SetStats(False)
h_train_bkg_prob.SetStats(False)
h_data_prob.SetStats(False)
# Print
h_train_sig_prob.Draw('HIST')
h_train_bkg_prob.Draw('HISTsame')
h_data_prob .Draw('HISTsame')
# Create legend
leg = ROOT.TLegend(0.6,0.75,0.9,0.9)
leg.AddEntry(h_train_sig_prob, 'Training Signal Events' , 'L')
leg.AddEntry(h_train_bkg_prob, 'Training Background Events', 'L')
leg.AddEntry(h_data_prob , 'Data Events' , 'L')
leg.SetLineColor(0)
leg.SetLineStyle(0)
leg.SetFillStyle(0)
leg.SetBorderSize(0)
leg.Draw('same')
# Save
c.SaveAs(out_path_run+'BDT_answer.pdf')
### Plot variable distributions
print(' *** Plotting training variable distributions ***')
out_path_var = out_path_run+'dist_vars/'
if not os.path.exists(out_path_var):
os.makedirs(out_path_var)
for var in D_CONFIGS[run]['fit_vars'] + ['bplus_PT'] + ['prob_'+run]:
# Initialise canvas
c_name = var+'_Distribution_'+run
c = ROOT.TCanvas(c_name, c_name, 600, 400)
c.cd()
# Select required quantity
a_plt_sig = df_train[var][df_train['prob_'+run] >= prob_threshold].as_matrix()
a_plt_bkg = df_train[var][df_train['prob_'+run] < prob_threshold].as_matrix()
# Scale DIRA and IPCHI2
i_str = ''
if (var=='bplus_DIRA_OWNPV'):
a_plt_sig = np.arccos(a_plt_sig)
a_plt_bkg = np.arccos(a_plt_bkg)
i_str = 'arccos '
if ('CHI2' in var):
a_plt_sig = np.log(a_plt_sig)
a_plt_bkg = np.log(a_plt_bkg)
i_str = 'log '
# Create and format histograms
x_max = max(max(a_plt_sig), max(a_plt_bkg))
x_min = min(min(a_plt_sig), min(a_plt_bkg))
h_plt_sig = ROOT.TH1F(c_name+'_Sig', c_name+'_Sig;'+i_str+var+';candidates', 100, x_min, x_max)
h_plt_bkg = ROOT.TH1F(c_name+'_Bkg', c_name+'_Bkg;'+i_str+var+';candidates', 100, x_min, x_max)
# Process full errors
h_plt_sig.Sumw2()
h_plt_bkg.Sumw2()
# Fill histograms
map(h_plt_sig.Fill, a_plt_sig)
map(h_plt_bkg.Fill, a_plt_bkg)
## Make it pretty
h_plt_sig.SetTitle(var+' Distribution '+run)
# Format for each case of x-axis
h_plt_sig.GetYaxis().SetTitleOffset(1.6)
y_max = 1.1*max(h_plt_sig.GetBinContent(h_plt_sig.GetMaximumBin()), h_plt_bkg.GetBinContent(h_plt_bkg.GetMaximumBin()))
h_plt_sig.GetYaxis().SetRangeUser(0, y_max)
h_plt_sig.GetXaxis().SetRangeUser(x_min, x_max)
# Format plotting style
h_plt_sig.SetLineColor(ROOT.kRed)
h_plt_sig.SetFillColorAlpha(ROOT .kRed - 10, 0.7)
h_plt_bkg.SetLineColor(ROOT.kBlue)
h_plt_bkg.SetFillColorAlpha(ROOT.kBlue - 10, 0.7)
# Remove stats boxes
h_plt_sig.SetStats(False)
h_plt_bkg.SetStats(False)
# Print
h_plt_sig.Draw('E0HIST')
h_plt_bkg.Draw('E0HISTsame')
# Create legend
leg = ROOT.TLegend(0.6,0.75,0.9,0.9)
leg.AddEntry(h_plt_sig, 'Training Events Identified as Signal', 'L')
leg.AddEntry(h_plt_bkg, 'Training Events Identified as Background', 'L')
leg.SetLineColor(0)
leg.SetLineStyle(0)
leg.SetFillStyle(0)
leg.SetBorderSize(0)
leg.Draw('same')
# Save
c.SaveAs(out_path_var+var+'.pdf')
### Plot comparison to MC data
out_path_mcp_data = out_path_run+'mcp_v_data/'
if not os.path.exists(out_path_mcp_data):
os.makedirs(out_path_mcp_data)
print(' *** Plotting comparison to psi(2S) MC ***')
df_data_comp = df_data[((df_data['mjpipi'] < 3696) & (df_data['mjpipi'] > 3676)) & ((df_data['scaledmass'] < 5299) & (df_data['scaledmass'] > 5259))]
df_side_comp = df_side[ (df_side['mjpipi'] < 3696) & (df_side['mjpipi'] > 3676)]
for var in D_CONFIGS[run]['fit_vars'] + ['bplus_PT'] + ['prob_'+run]:
# Initialise canvas
c_name = var+'_MC_#psi(2S)_Comparison_'+run
c = ROOT.TCanvas(c_name, c_name, 600, 400)
c.cd()
# Select required quantity
a_mc_p = df_mc_p[var].as_matrix()
a_data_comp = df_data_comp[var].as_matrix()
a_side_comp = df_side_comp[var].as_matrix()
# Scale DIRA and IPCHI2
i_str = ''
if (var=='bplus_DIRA_OWNPV'):
a_mc_p = np.arccos(a_mc_p)
a_data_comp = np.arccos(a_data_comp)
a_side_comp = np.arccos(a_side_comp)
i_str = 'arccos '
#if ('CHI2' in var):
# a_mc_p = np.log(a_mc_p)
# a_data_comp = np.log(a_data_comp)
# a_side_comp = np.log(a_side_comp)
# i_str = 'log '
# Create and format histograms
x_max = max(max(a_mc_p), max(a_data_comp))
x_min = min(min(a_mc_p), min(a_data_comp))
h_mc_p = ROOT.TH1F(c_name+'_mc_#psi', c_name+'_mc_#psi;'+i_str+var+';candidates', 100, x_min, x_max)
h_comp = ROOT.TH1F(c_name+'_data' , c_name+'_data;'+i_str+var+';candidates' , 100, x_min, x_max)
h_side = ROOT.TH1F(c_name+'_side' , c_name+'_side;'+i_str+var+';candidates' , 100, x_min, x_max)
# Process errors correctly
h_mc_p.Sumw2()
h_comp.Sumw2()
h_side.Sumw2()
# Fill histograms
map(h_mc_p.Fill, a_mc_p)
map(h_comp.Fill, a_data_comp)
map(h_side.Fill, a_side_comp)
# Background reduce
h_comp.Add(h_side, -1)
# Normalise
h_mc_p.Scale(1./h_mc_p.Integral())
h_comp.Scale(1./h_comp.Integral())
h_side.Scale(1./h_side.Integral())
## Make it pretty
h_mc_p.SetTitle(var+' Data vs MC J(2S) Distribution '+run)
# Format for each case of x-axis
h_mc_p.GetYaxis().SetTitleOffset(1.6)
y_max = 1.1*max((h_mc_p.GetBinContent(h_mc_p.GetMaximumBin()), h_comp.GetBinContent(h_comp.GetMaximumBin()), h_side.GetBinContent(h_side.GetMaximumBin())))
y_min = 0.9*min((h_mc_p.GetBinContent(h_mc_p.GetMinimumBin()), h_comp.GetBinContent(h_comp.GetMinimumBin()), h_side.GetBinContent(h_side.GetMinimumBin())))
h_mc_p.GetYaxis().SetRangeUser(y_min, y_max)
# Format plotting style
h_mc_p.SetLineColor(ROOT.kRed)
h_mc_p.SetFillColorAlpha(ROOT .kRed - 10, 0.7)
h_comp.SetLineColor(ROOT.kBlue)
h_comp.SetFillColorAlpha(ROOT .kBlue - 10, 0.7)
h_side.SetLineColor(ROOT.kGreen)
h_side.SetFillColorAlpha(ROOT.kGreen - 10, 0.7)
# Remove stats boxes
h_mc_p.SetStats(False)
h_comp.SetStats(False)
h_side.SetStats(False)
# Print
h_mc_p.Draw('E0HIST')
h_comp.Draw('E0HISTsame')
h_side.Draw('E0HISTsame')
# Create legend
leg = ROOT.TLegend(0.6,0.75,0.9,0.9)
leg.AddEntry(h_mc_p, '#psi(2S) Monte-Carlo', 'L')
leg.AddEntry(h_comp, 'Background Reduced Data in #psi(2S) Region', 'L')
leg.AddEntry(h_side, 'Background Data in #psi(2S) Region', 'L')
leg.SetLineColor(0)
leg.SetLineStyle(0)
leg.SetFillStyle(0)
leg.SetBorderSize(0)
leg.Draw('same')
# Save
c.SaveAs(out_path_mcp_data+var+'.pdf')
### Plot comparison to MC data
out_path_mcx_data = out_path_run+'mcx_v_data/'
if not os.path.exists(out_path_mcx_data):
os.makedirs(out_path_mcx_data)
print(' *** Plotting comparison to X(3872) MC ***')
df_data_comp = df_data[((df_data['mjpipi'] < 3882) & (df_data['mjpipi'] > 3862)) & ((df_data['scaledmass'] < 5299) & (df_data['scaledmass'] > 5259))]
df_side_comp = df_side[ (df_side['mjpipi'] < 3882) & (df_side['mjpipi'] > 3862)]
for var in D_CONFIGS[run]['fit_vars'] + ['bplus_PT'] +['prob_'+run]:
# Initialise canvas
c_name = var+'_MC_X(3872)_Comparison_'+run
c = ROOT.TCanvas(c_name, c_name, 600, 400)
c.cd()
# Select required quantity
a_mc_p = df_mc_x[var].as_matrix()
a_data_comp = df_data_comp[var].as_matrix()
a_side_comp = df_side_comp[var].as_matrix()
# Scale DIRA and IPCHI2
i_str = ''
if (var=='bplus_DIRA_OWNPV'):
a_mc_p = np.arccos(a_mc_p)
a_data_comp = np.arccos(a_data_comp)
a_side_comp = np.arccos(a_side_comp)
i_str = 'arccos '
if ('CHI2' in var):
a_mc_p = np.log(a_mc_p)
a_data_comp = np.log(a_data_comp)
a_side_comp = np.log(a_side_comp)
i_str = 'log '
# Create and format histograms
x_max = max(max(a_mc_p), max(a_data_comp))
x_min = min(min(a_mc_p), min(a_data_comp))
h_mc_p = ROOT.TH1F(c_name+'_mc_#psi', c_name+'_mc_#psi;'+i_str+var+';candidates', 100, x_min, x_max)
h_comp = ROOT.TH1F(c_name+'_data' , c_name+'_data;'+i_str+var+';candidates' , 100, x_min, x_max)
h_side = ROOT.TH1F(c_name+'_side' , c_name+'_side;'+i_str+var+';candidates' , 100, x_min, x_max)
# Process errors correctly
h_mc_p.Sumw2()
h_comp.Sumw2()
h_side.Sumw2()
# Fill histograms
map(h_mc_p.Fill, a_mc_p)
map(h_comp.Fill, a_data_comp)
map(h_side.Fill, a_side_comp)
# Background reduce
h_comp.Add(h_side, -1)
# Normalise
h_mc_p.Scale(1./h_mc_p.Integral())
h_comp.Scale(1./h_comp.Integral())
h_side.Scale(1./h_side.Integral())
## Make it pretty
h_mc_p.SetTitle(var+' Data vs MC X(3872) Distribution '+run)
# Format for each case of x-axis
h_mc_p.GetYaxis().SetTitleOffset(1.6)
y_max = 1.1*max((h_mc_p.GetBinContent(h_mc_p.GetMaximumBin()), h_comp.GetBinContent(h_comp.GetMaximumBin()), h_side.GetBinContent(h_side.GetMaximumBin())))
y_min = 0.9*min((h_mc_p.GetBinContent(h_mc_p.GetMinimumBin()), h_comp.GetBinContent(h_comp.GetMinimumBin()), h_side.GetBinContent(h_side.GetMinimumBin())))
h_mc_p.GetYaxis().SetRangeUser(y_min, y_max)
# Format plotting style
h_mc_p.SetLineColor(ROOT.kRed)
h_mc_p.SetFillColorAlpha(ROOT .kRed - 10, 0.7)
h_comp.SetLineColor(ROOT.kBlue)
h_comp.SetFillColorAlpha(ROOT .kBlue - 10, 0.7)
h_side.SetLineColor(ROOT.kGreen)
h_side.SetFillColorAlpha(ROOT.kGreen - 10, 0.7)
# Remove stats boxes
h_mc_p.SetStats(False)
h_comp.SetStats(False)
h_side.SetStats(False)
# Print
h_mc_p.Draw('E0HIST')
h_comp.Draw('E0HISTsame')
h_side.Draw('E0HISTsame')
# Create legend
leg = ROOT.TLegend(0.6,0.75,0.9,0.9)
leg.AddEntry(h_mc_p, 'X(3872) Monte-Carlo', 'L')
leg.AddEntry(h_comp, 'Background Reduced Data in X(3872) Region', 'L')
leg.AddEntry(h_side, 'Background Data in X(3872) Region', 'L')
leg.SetLineColor(0)
leg.SetLineStyle(0)
leg.SetFillStyle(0)
leg.SetBorderSize(0)
leg.Draw('same')
# Save
c.SaveAs(out_path_mcx_data+var+'.pdf')
### Plot comparison of MC data
out_path_mc_mc = out_path_run+'mcp_v_mcx/'
if not os.path.exists(out_path_mc_mc):
os.makedirs(out_path_mc_mc)
print(' *** Plotting comparison of psi(2S) MC and X(3972) MC ***')
for var in D_CONFIGS[run]['fit_vars'] + ['bplus_PT'] + ['prob_'+run]:
# Initialise canvas
c_name = var+'_MC_Comparison_'+run
c = ROOT.TCanvas(c_name, c_name, 600, 400)
c.cd()
# Select required quantity
a_mc_p = df_mc_p[var].as_matrix()
a_mc_x = df_mc_x[var].as_matrix()
# Scale DIRA and IPCHI2
i_str = ''
if (var=='bplus_DIRA_OWNPV'):
a_mc_p = np.arccos(a_mc_p)
a_mc_x = np.arccos(a_mc_x)
i_str = 'arccos '
if ('CHI2' in var):
a_mc_p = np.log(a_mc_p)
a_mc_x = np.log(a_mc_x)
i_str = 'log '
# Create and format histograms
x_max = max(max(a_mc_p), max(a_mc_x))
x_min = min(min(a_mc_p), min(a_mc_x))
h_mc_p = ROOT.TH1F(c_name+'_mc_#psi', c_name+'_mc_#psi;'+i_str+var+';candidates', 100, x_min, x_max)
h_mc_x = ROOT.TH1F(c_name+'_mc_X' , c_name+'_mc_X;'+i_str+var+';candidates' , 100, x_min, x_max)
# Process errors correctly
h_mc_p.Sumw2()
h_mc_x.Sumw2()
# Fill histograms
map(h_mc_p.Fill, a_mc_p)
map(h_mc_x.Fill, a_mc_x)
# Normalise
h_mc_p.Scale(1./h_mc_p.Integral())
h_mc_x.Scale(1./h_mc_x.Integral())
## Make it pretty
h_mc_p.SetTitle(var+' MC X(3872) vs MC #psi(2S) Distribution '+run)
# Format for each case of x-axis
h_mc_p.GetYaxis().SetTitleOffset(1.6)
y_max = 1.1*max(h_mc_p.GetBinContent(h_mc_p.GetMaximumBin()), h_mc_x.GetBinContent(h_mc_x.GetMaximumBin()))
y_min = 0.9*min(h_mc_p.GetBinContent(h_mc_p.GetMinimumBin()), h_mc_x.GetBinContent(h_mc_x.GetMinimumBin()))
h_mc_p.GetYaxis().SetRangeUser(y_min, y_max)
# Format plotting style
h_mc_p.SetLineColor(ROOT.kRed)
h_mc_p.SetFillColorAlpha(ROOT .kRed - 10, 0.7)
h_mc_x.SetLineColor(ROOT.kBlue)
h_mc_x.SetFillColorAlpha(ROOT.kBlue - 10, 0.7)
# Remove stats boxes
h_mc_p.SetStats(False)
h_mc_x.SetStats(False)
# Print
h_mc_p.Draw('E0HIST')
h_mc_x.Draw('E0HISTsame')
# Create legend
leg = ROOT.TLegend(0.6,0.75,0.9,0.9)
leg.AddEntry(h_mc_p, '#psi(2S) Monte-Carlo', 'L')
leg.AddEntry(h_mc_x, 'X(3872) Monte-Carlo', 'L')
leg.SetLineColor(0)
leg.SetLineStyle(0)
leg.SetFillStyle(0)
leg.SetBorderSize(0)
leg.Draw('same')
# Save
c.SaveAs(out_path_mc_mc+var+'.pdf')
## Perform fit to XGB cut data
# Filter dataframes
a_cut_mc = df_train[df_train['class'] == 1]['scaledmass'].as_matrix()
a_cut_data = df_data[df_data['class'] == 1]['scaledmass'].as_matrix()
# Fit
d_cut_fit = fit_doubleCB(a_cut_mc, a_cut_data, out_path_run, s_info='cut_data_plot')
# Store params
d_run_info[run]['cut_fit_params'] = d_cut_fit
# Fit in X region only
a_mc_x = df_train[df_train['cat']=='mc_x']['scaledmass'].as_matrix()
a_data_x = df_data[(df_data['class'] == 1) & ((df_data['mjpipi'] > 3862) & (df_data['mjpipi'] < 3882))]['scaledmass'].as_matrix()
d_sig_est = fit_doubleCB(a_mc_x, a_data_x, out_path_run, s_info='x_signal_yield_est')
d_run_info[run]['x_reg_fit_params'] = d_sig_est
print( '*** Estimated fitted signal efficiency: {:.3f} ***'.format(float(d_cut_fit['data_sig_yield'])/d_sig_est_alldata['data_sig_yield']))
print('*** Plotting ROC curve ***')
### Plot ROC curve
fig = plt.figure()
for run in list(D_CONFIGS.keys()):
plt.plot(d_roc_plot[run]['bgr_rejs'], d_roc_plot[run]['sig_effs'], label=run)
plt.legend(loc=3)
plt.ylabel("Background Rejection")
plt.xlabel("Signal Efficiency")
plt.xlim(0.,1.)
plt.ylim(0.,1.)
plt.title("ROC Curve")
plt.tight_layout(pad=2.0)
fig.savefig(out_path+'ROC_curve.pdf')
plt.close()
print('*** Dumping run information ***')
with open(out_path + args.out_dict, 'w') as outfile:
yaml.dump(d_run_info, outfile, default_flow_style=False)
with open(out_path + 'roc_plot.yml', 'w') as outfile:
yaml.dump(d_roc_plot, outfile, default_flow_style=False)
if __name__ == "__main__":
parser = argparse.ArgumentParser(description = "--add description--")
parser.add_argument("data_dir" , default = None , help = "directory containing trained root files")
parser.add_argument("--out_path" , "-o", default = 'results/' , help = "directory for storing results")
parser.add_argument("--config" , "-c", default = 'config/xgboost.yml', help = "config file")
parser.add_argument("--out_dict" , "-q", default = 'run_info.yml' , help = "dictionary summarising run information")
parser.add_argument("--tree_name", "-t", default = "DecayTree" , help = "input tree name")
parser.add_argument("--find_s0" , "-s", action = 'store_true' , help = "if specified, a fit will be made to estimate s0 for X(3823)")
parser.add_argument("--bck_cut" , "-b", action = 'store_true' , help = "if specified, the optimal cut will be where background rejection reaches 99%")
parser.add_argument("--opt_cut" , "-r", default = None , help = "if specified, skip optimisation and take this as the cut factor")
args = parser.parse_args()
main(args)