In [1]:

    

%pylab inline


    

    




Populating the interactive namespace from numpy and matplotlib

In [39]:

    

%run additional.ipynb


    

In [3]:

    

pandas.set_option('display.max_colwidth', 120)


    

In [4]:

    

bck_train_mode_name = 30000000
sig_train_modes_names = list(set(empty_events.keys()) - {bck_train_mode_name})
sig_train_files = ['mod_{}.csv'.format(name) for name in sig_train_modes_names]
bck_train_files = 'mod_30000000.csv'
folder = "datasets/prepared_hlt_body/"


    

In [5]:

    

# concat all signal data
if not os.path.exists(folder + 'signal_hlt1.csv'):
    concat_files(folder, sig_train_files, os.path.join(folder , 'signal_hlt1.csv'))


    

In [6]:

    

signal_data = pandas.read_csv(os.path.join(folder, 'signal_hlt1.csv'), sep='\t')
bck_data = pandas.read_csv(os.path.join(folder, bck_train_files), sep='\t')


    

In [7]:

    

signal_data.columns


    

    
Out[7]:





Index([u'unique', u'mode', u'event_number', u'sv_number', u'pass_2body', u'pass_nbody', u'signal', u'sumpt', u'm', u'mcor', u'ipchi2', u'chi2', u'sumipchi2', u'fdr', u'nlt16', u'minpt', u'eta', u'pt', u'nmu', u'n', u'fdchi2', u'maxtchi2', u'ngood', u'nmu1', u'mupt', u'n1trk', u'sig', u'idx'], dtype='object')

In [8]:

    

print 'Signal', statistic_length(signal_data)
print 'Bck', statistic_length(bck_data)


    

    




Signal {'SVR': 28773384, 'Events': 560508}
Bck {'SVR': 408147, 'Events': 38035}

In [9]:

    

total_bck_events = statistic_length(bck_data)['Events'] + empty_events[bck_train_mode_name]
total_signal_events_by_mode = dict()
for mode in sig_train_modes_names:
    total_signal_events_by_mode[mode] = statistic_length(signal_data[signal_data['mode'] == mode])['Events'] + empty_events[mode]


    

In [10]:

    

print 'Bck:', total_bck_events
'Signal:', total_signal_events_by_mode


    

    




Bck: 111306






    
Out[10]:





('Signal:',
 {11102003: 22317,
  11104121: 21960,
  11114001: 61102,
  11114101: 36257,
  11124001: 0,
  11296013: 17149,
  11536011: 13893,
  11874004: 27212,
  11874042: 3942,
  12103035: 25250,
  12103121: 22180,
  12165106: 0,
  12265042: 0,
  12873002: 9480,
  12873432: 0,
  12875500: 0,
  13102201: 0,
  13102412: 31938,
  13104012: 0,
  13112001: 34736,
  13144001: 54266,
  13144020: 39278,
  13246001: 37313,
  13264021: 25165,
  13512010: 43379,
  13774002: 12589,
  15164001: 7130,
  15512011: 27799,
  20000000: 0,
  21263002: 8871,
  23103110: 0,
  25103000: 0,
  25103110: 0,
  27163002: 848,
  27163400: 380,
  27265001: 871,
  27265101: 2192})

In [11]:

    

variables_base = ["mcor", "chi2", "pt", "fdchi2", "minpt", "nlt16"]
variables_mcor = ["chi2", "pt", "fdchi2", "minpt", "nlt16"]
variables_additional = ["m", "fdr", "sumpt", "sumipchi2", "eta"]
variables_new = ['chi2', 'minpt', 'sumpt', 'fdchi2', 'nlt16']
variables_new_minpt = ['chi2', 'sumpt', 'fdchi2', 'nlt16']


    

In [12]:

    

# hlt1 2body selection
signal_data = signal_data[signal_data['pass_2body'] == 1]
bck_data = bck_data[bck_data['pass_2body'] == 1]


    

In [13]:

    

print 'Signal', statistic_length(signal_data)
print 'Bck', statistic_length(bck_data)


    

    




Signal {'SVR': 1154688, 'Events': 451748}
Bck {'SVR': 21861, 'Events': 12961}

In [14]:

    

total_signal_events_by_mode_presel = dict()
for mode in sig_train_modes_names:
    total_signal_events_by_mode_presel[mode] = statistic_length(signal_data[signal_data['mode'] == mode])['Events']
total_bck_events_presel = statistic_length(bck_data)['Events']


    

In [15]:

    

print 'Bck:', total_bck_events_presel
'Signal:', total_signal_events_by_mode_presel


    

    




Bck: 12961






    
Out[15]:





('Signal:',
 {11102003: 17238,
  11104121: 13873,
  11114001: 53794,
  11114101: 24563,
  11124001: 0,
  11296013: 16034,
  11536011: 12056,
  11874004: 22509,
  11874042: 3178,
  12103035: 21536,
  12103121: 2299,
  12165106: 0,
  12265042: 0,
  12873002: 7390,
  12873432: 0,
  12875500: 0,
  13102201: 0,
  13102412: 18919,
  13104012: 0,
  13112001: 28896,
  13144001: 48545,
  13144020: 34670,
  13246001: 33963,
  13264021: 22178,
  13512010: 28333,
  13774002: 10332,
  15164001: 6362,
  15512011: 17642,
  20000000: 0,
  21263002: 5808,
  23103110: 0,
  25103000: 0,
  25103110: 0,
  27163002: 460,
  27163400: 135,
  27265001: 406,
  27265101: 629})

In [16]:

    

signal_data.head()


    

    
Out[16]:






  

    

      

      
unique
      
mode
      
event_number
      
sv_number
      
pass_2body
      
pass_nbody
      
signal
      
sumpt
      
m
      
mcor
      
...
      
nmu
      
n
      
fdchi2
      
maxtchi2
      
ngood
      
nmu1
      
mupt
      
n1trk
      
sig
      
idx
    
  
  

    

      
0 
      
 13144001_0
      
 13144001
      
 0
      
 0
      
 1
      
 1
      
 1
      
 2919.64
      
 1723.000
      
 4091.68
      
...
      
 1
      
 2
      
  4673.5800
      
 1.24344
      
 2
      
 1
      
 1846.81
      
 2
      
 1
      
 0
    
    

      
3 
      
 13144001_0
      
 13144001
      
 0
      
 3
      
 1
      
 1
      
 1
      
 2463.47
      
  337.013
      
 1578.69
      
...
      
 0
      
 2
      
    75.1089
      
 1.09587
      
 2
      
 0
      
   -1.00
      
 2
      
 1
      
 3
    
    

      
4 
      
 13144001_0
      
 13144001
      
 0
      
 4
      
 1
      
 1
      
 1
      
 3237.45
      
 2165.520
      
 4253.02
      
...
      
 1
      
 2
      
  5344.6800
      
 1.24344
      
 2
      
 1
      
 1846.81
      
 2
      
 1
      
 4
    
    

      
38
      
 13144001_1
      
 13144001
      
 1
      
 3
      
 1
      
 1
      
 1
      
 4420.88
      
 1645.150
      
 4260.60
      
...
      
 1
      
 2
      
  7155.3000
      
 1.30690
      
 2
      
 1
      
 3761.62
      
 1
      
 1
      
 3
    
    

      
39
      
 13144001_1
      
 13144001
      
 1
      
 4
      
 1
      
 1
      
 1
      
 4498.91
      
 1911.790
      
 4473.97
      
...
      
 1
      
 2
      
 19735.2000
      
 1.15761
      
 2
      
 1
      
 3761.62
      
 1
      
 1
      
 4
    
  

5 rows × 28 columns

In [17]:

    

ds_train_signal, ds_train_bck, ds_test_signal, ds_test_bck = prepare_data(signal_data, bck_data, 'unique')


    

In [18]:

    

print 'Signal', statistic_length(ds_train_signal)
print 'Bck', statistic_length(ds_train_bck)


    

    




Signal {'SVR': 577894, 'Events': 225874}
Bck {'SVR': 10853, 'Events': 6480}

In [19]:

    

train = pandas.concat([ds_train_bck, ds_train_signal])


    

In [20]:

    

print 'Signal', statistic_length(ds_test_signal)
print 'Bck', statistic_length(ds_test_bck)


    

    




Signal {'SVR': 576794, 'Events': 225874}
Bck {'SVR': 11008, 'Events': 6481}

In [21]:

    

test = pandas.concat([ds_test_bck, ds_test_signal])


    

In [22]:

    

total_test_bck_events = (total_bck_events - total_bck_events_presel) // 2 + statistic_length(ds_test_bck)['Events']
total_test_signal_events = dict()
for mode in sig_train_modes_names:
    total_not_passed_signal = total_signal_events_by_mode[mode] - total_signal_events_by_mode_presel[mode]
    total_test_signal_events[mode] = total_not_passed_signal // 2 + \
        statistic_length(ds_test_signal[ds_test_signal['mode'] == mode])['Events']


    

In [23]:

    

print 'Bck total test events:', total_test_bck_events
'Signal total test events:', total_test_signal_events


    

    




Bck total test events: 55653






    
Out[23]:





('Signal total test events:',
 {11102003: 11170,
  11104121: 10965,
  11114001: 30653,
  11114101: 18218,
  11124001: 0,
  11296013: 8479,
  11536011: 6991,
  11874004: 13532,
  11874042: 2001,
  12103035: 12636,
  12103121: 11092,
  12165106: 0,
  12265042: 0,
  12873002: 4778,
  12873432: 0,
  12875500: 0,
  13102201: 0,
  13102412: 15943,
  13104012: 0,
  13112001: 17346,
  13144001: 27102,
  13144020: 19656,
  13246001: 18607,
  13264021: 12544,
  13512010: 21717,
  13774002: 6320,
  15164001: 3553,
  15512011: 13794,
  20000000: 0,
  21263002: 4472,
  23103110: 0,
  25103000: 0,
  25103110: 0,
  27163002: 429,
  27163400: 192,
  27265001: 452,
  27265101: 1099})

In [24]:

    

# cut on mcor
train_cut = train[train['mcor'] <= 10e3]


    

In [25]:

    

for var in variables_base + variables_additional:
    hist(train[train.signal == 1][var].values, color='b', bins=60, histtype='step', normed=True)
    hist(train[train.signal == 0][var].values, color='r', bins=60, histtype='step', normed=True)
    title(var)
    show()


    

In [26]:

    

import cPickle
if os.path.exists('models/hlt1_body2.pkl'):
    with open('models/hlt1_body2.pkl', 'r') as file_mn:
        estimators = cPickle.load(file_mn)


    

In [27]:

    

from rep_ef.estimators import EventFilterClassifier


    

In [27]:

    

estimators['MN: mcor cut, mcor var'] = EventFilterClassifier(features=variables_base,
                                                            dataset_name='hlt1_2{random}', 
                                                            iterations=5000, sync=False)
estimators['MN: mcor cut, mcor var'].fit(train_cut, train_cut['signal'])


    

    
Out[27]:





EventFilterClassifier(auto_stop=None, baseline=None, command_line_params=None,
           connection='default', dataset_name='hlt1_2877378',
           dump_filename=None,
           features=['mcor', 'chi2', 'pt', 'fdchi2', 'minpt', 'nlt16'],
           features_sample_rate_per_iteration=1.0, formula_name='EF',
           intervals=64, iterations=5000, max_features_per_iteration=6,
           regularization=0.01, sync=False, training_fraction=0.5)

In [26]:

    

estimators['MN: mcor cut, mcor var'].get_feature_importances().sort('effect')


    

    
Out[26]:






  

    

      

      
effect
      
efficiency
      
information
    
  
  

    

      
chi2
      
 0.339338
      
 0.077
      
 1.000000
    
    

      
minpt
      
 0.391074
      
 0.092
      
 0.955473
    
    

      
nlt16
      
 0.603319
      
 1.000
      
 0.136052
    
    

      
pt
      
 0.604010
      
 0.145
      
 0.939794
    
    

      
fdchi2
      
 0.903096
      
 0.222
      
 0.916252
    
    

      
mcor
      
 1.000000
      
 0.226
      
 0.997089
    
  

In [28]:

    

estimators['MN'] = EventFilterClassifier(features=variables_mcor,
                                         dataset_name='hlt1_2{random}', 
                                         iterations=5000, sync=False)
estimators['MN'].fit(train, train['signal'])


    

    
Out[28]:





EventFilterClassifier(auto_stop=None, baseline=None, command_line_params=None,
           connection='default', dataset_name='hlt1_2601967',
           dump_filename=None,
           features=['chi2', 'pt', 'fdchi2', 'minpt', 'nlt16'],
           features_sample_rate_per_iteration=1.0, formula_name='EF',
           intervals=64, iterations=5000, max_features_per_iteration=6,
           regularization=0.01, sync=False, training_fraction=0.5)

In [27]:

    

estimators['MN'].get_feature_importances().sort('effect')


    

    
Out[27]:






  

    

      

      
effect
      
efficiency
      
information
    
  
  

    

      
chi2
      
 0.365566
      
 0.093
      
 1.000000
    
    

      
minpt
      
 0.392008
      
 0.105
      
 0.950177
    
    

      
nlt16
      
 0.548953
      
 1.000
      
 0.140354
    
    

      
pt
      
 0.638026
      
 0.173
      
 0.944490
    
    

      
fdchi2
      
 1.000000
      
 0.279
      
 0.917606
    
  

In [29]:

    

estimators['additional features'] = EventFilterClassifier(features=variables_base + variables_additional,
                                                           dataset_name='hlt1_2{random}', 
                                                           iterations=5000, sync=False)
estimators['additional features'].fit(train_cut, train_cut['signal'])


    

    
Out[29]:





EventFilterClassifier(auto_stop=None, baseline=None, command_line_params=None,
           connection='default', dataset_name='hlt1_2327529',
           dump_filename=None,
           features=['mcor', 'chi2', 'pt', 'fdchi2', 'minpt', 'nlt16', 'm', 'fdr', 'sumpt', 'sumipchi2', 'eta'],
           features_sample_rate_per_iteration=1.0, formula_name='EF',
           intervals=64, iterations=5000, max_features_per_iteration=6,
           regularization=0.01, sync=False, training_fraction=0.5)

In [28]:

    

estimators['additional features'].get_feature_importances().sort('effect')


    

    
Out[28]:






  

    

      

      
effect
      
efficiency
      
information
    
  
  

    

      
pt
      
 0.167901
      
 0.034
      
 0.922125
    
    

      
chi2
      
 0.309723
      
 0.059
      
 0.981200
    
    

      
sumipchi2
      
 0.324913
      
 0.068
      
 0.899348
    
    

      
minpt
      
 0.372659
      
 0.074
      
 0.937510
    
    

      
eta
      
 0.374129
      
 0.070
      
 1.000000
    
    

      
fdr
      
 0.413328
      
 0.086
      
 0.901855
    
    

      
m
      
 0.485192
      
 0.093
      
 0.982099
    
    

      
sumpt
      
 0.617051
      
 0.125
      
 0.921341
    
    

      
fdchi2
      
 0.666376
      
 0.139
      
 0.899026
    
    

      
nlt16
      
 0.712979
      
 1.000
      
 0.133494
    
    

      
mcor
      
 1.000000
      
 0.191
      
 0.978344
    
  

In [31]:

    

estimators['MN: pt->sumpt'] = EventFilterClassifier(features=variables_new, 
                                                    dataset_name='hlt1_2{random}', 
                                                    iterations=5000, sync=False)
estimators['MN: pt->sumpt'].fit(train, train['signal'])


    

    
Out[31]:





EventFilterClassifier(auto_stop=None, baseline=None, command_line_params=None,
           connection='default', dataset_name='hlt1_2983847',
           dump_filename=None,
           features=['chi2', 'minpt', 'sumpt', 'fdchi2', 'nlt16'],
           features_sample_rate_per_iteration=1.0, formula_name='EF',
           intervals=64, iterations=5000, max_features_per_iteration=6,
           regularization=0.01, sync=False, training_fraction=0.5)

In [29]:

    

estimators['MN: pt->sumpt'].get_feature_importances().sort('effect')


    

    
Out[29]:






  

    

      

      
effect
      
efficiency
      
information
    
  
  

    

      
minpt
      
 0.366321
      
 0.092
      
 0.950177
    
    

      
chi2
      
 0.379072
      
 0.091
      
 1.000000
    
    

      
nlt16
      
 0.587083
      
 1.000
      
 0.140354
    
    

      
sumpt
      
 0.626017
      
 0.159
      
 0.942135
    
    

      
fdchi2
      
 1.000000
      
 0.261
      
 0.917606
    
  

In [33]:

    

estimators['MN: pt->sumpt, remove minpt'] = EventFilterClassifier(features=variables_new_minpt,
                                                                  dataset_name='hlt1_2{random}', 
                                                                  iterations=5000, sync=False)
estimators['MN: pt->sumpt, remove minpt'].fit(train, train['signal'])


    

    
Out[33]:





EventFilterClassifier(auto_stop=None, baseline=None, command_line_params=None,
           connection='default', dataset_name='hlt1_2498070',
           dump_filename=None,
           features=['chi2', 'sumpt', 'fdchi2', 'nlt16'],
           features_sample_rate_per_iteration=1.0, formula_name='EF',
           intervals=64, iterations=5000, max_features_per_iteration=6,
           regularization=0.01, sync=False, training_fraction=0.5)

In [34]:

    

estimators['MN: pt->sumpt, remove minpt'].get_feature_importances().sort('effect')


    

    
Out[34]:






  

    

      

      
effect
      
efficiency
      
information
    
  
  

    

      
chi2
      
 0.362978
      
 0.089
      
 1.000000
    
    

      
nlt16
      
 0.572225
      
 1.000
      
 0.140354
    
    

      
sumpt
      
 0.671271
      
 0.175
      
 0.942135
    
    

      
fdchi2
      
 1.000000
      
 0.267
      
 0.917606
    
  

In [35]:

    

borders = {
'chi2': [1,2.5,5,7.5,10,100], 
'sumpt': [3000,4000,5000,6000,7500,9000,12e3,23e3,50e3],  
'fdchi2': [33,125,350,780,1800,5000,10000], 
'minpt': [350,500,750,1500,3000,5000], 
'nlt16': [0.5]
}


    

In [36]:

    

borders_minpt = {
'chi2': [1,2.5,5,7.5,10,100],  
'sumpt': [3000,4000,5000,6000,7500,9000,12e3,23e3,50e3],  
'fdchi2': [33,125,350,780,1800,5000,10000],  
'nlt16': [0.5]
}


    

In [37]:

    

estimators['MN BBDT: pt->sumpt'] = EventFilterClassifier(features=variables_new,
                                                        dataset_name='hlt1_2{random}', 
                                                        iterations=5000, sync=False, 
                                                        intervals=borders)
estimators['MN BBDT: pt->sumpt'].fit(train, train['signal'])


    

    
Out[37]:





EventFilterClassifier(auto_stop=None, baseline=None, command_line_params=None,
           connection='default', dataset_name='hlt1_2792572',
           dump_filename=None,
           features=['chi2', 'minpt', 'sumpt', 'fdchi2', 'nlt16'],
           features_sample_rate_per_iteration=1.0, formula_name='EF',
           intervals={'chi2': [1, 2.5, 5, 7.5, 10, 100], 'sumpt': [3000, 4000, 5000, 6000, 7500, 9000, 12000.0, 23000.0, 50000.0], 'nlt16': [0.5], 'fdchi2': [33, 125, 350, 780, 1800, 5000, 10000], 'minpt': [350, 500, 750, 1500, 3000, 5000]},
           iterations=5000, max_features_per_iteration=6,
           regularization=0.01, sync=False, training_fraction=0.5)

In [45]:

    

estimators['MN BBDT: pt->sumpt'].get_feature_importances().sort('effect')


    

    
Out[45]:






  

    

      

      
effect
      
efficiency
      
information
    
  
  

    

      
minpt
      
 0.244235
      
 0.126
      
 0.633747
    
    

      
chi2
      
 0.319598
      
 0.182
      
 0.573541
    
    

      
sumpt
      
 0.616134
      
 0.202
      
 1.000000
    
    

      
nlt16
      
 0.818983
      
 1.000
      
 0.268050
    
    

      
fdchi2
      
 1.000000
      
 0.330
      
 0.990346
    
  

In [38]:

    

estimators['MN BBDT: pt->sumpt, remove minpt'] = EventFilterClassifier(features=variables_new_minpt,
                                                                        dataset_name='hlt1_2{random}', 
                                                                        iterations=5000, sync=False, 
                                                                        intervals=borders_minpt)
estimators['MN BBDT: pt->sumpt, remove minpt'].fit(train, train['signal'])


    

    
Out[38]:





EventFilterClassifier(auto_stop=None, baseline=None, command_line_params=None,
           connection='default', dataset_name='hlt1_279902',
           dump_filename=None,
           features=['chi2', 'sumpt', 'fdchi2', 'nlt16'],
           features_sample_rate_per_iteration=1.0, formula_name='EF',
           intervals={'chi2': [1, 2.5, 5, 7.5, 10, 100], 'sumpt': [3000, 4000, 5000, 6000, 7500, 9000, 12000.0, 23000.0, 50000.0], 'nlt16': [0.5], 'fdchi2': [33, 125, 350, 780, 1800, 5000, 10000]},
           iterations=5000, max_features_per_iteration=6,
           regularization=0.01, sync=False, training_fraction=0.5)

In [46]:

    

estimators['MN BBDT: pt->sumpt, remove minpt'].get_feature_importances().sort('effect')


    

    
Out[46]:






  

    

      

      
effect
      
efficiency
      
information
    
  
  

    

      
chi2
      
 0.248029
      
 0.191
      
 0.573541
    
    

      
sumpt
      
 0.557311
      
 0.246
      
 1.000000
    
    

      
nlt16
      
 0.606217
      
 1.000
      
 0.268050
    
    

      
fdchi2
      
 1.000000
      
 0.446
      
 0.990346
    
  

In [28]:

    

from rep.metaml import FoldingClassifier
from rep.estimators import SklearnClassifier
from sklearn.ensemble import RandomForestClassifier


    

In [30]:

    

forest_base_partial = SklearnClassifier(RandomForestClassifier(n_estimators=300, min_samples_leaf=500, max_depth=7,
                                                               max_features=4))
forest_folding_top = FoldingClassifier(base_estimator=forest_base_partial, random_state=11, 
                                       features=variables_new_minpt, ipc_profile='ssh-ipy')
forest_folding_top.fit(train, train['signal'])


    

    
Out[30]:





FoldingClassifier(base_estimator=SklearnClassifier(clf=RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
            max_depth=7, max_features=4, max_leaf_nodes=None,
            min_samples_leaf=500, min_samples_split=2,
            min_weight_fraction_leaf=0.0, n_estimators=300, n_jobs=1,
            oob_score=False, random_state=None, verbose=0,
            warm_start=False),
         features=None),
         features=['chi2', 'sumpt', 'fdchi2', 'nlt16'],
         ipc_profile='ssh-ipy', n_folds=2, random_state=11)

In [47]:

    

for rank in range(1, 3):
    good_events = get_best_svr(train, forest_folding_top, rank)
    ef_good = EventFilterClassifier(features=variables_base,
                                    dataset_name='hlt1_2{random}',
                                    iterations=5000, sync=False)
    ef_good.fit(good_events, good_events['signal'])
    estimators['top-{} forest preselection, use mcor'.format(rank)] = ef_good


    

    




KFold prediction using folds column
352627

In [32]:

    

for rank in range(1, 3):
    good_events = get_best_svr(train, forest_folding_top, rank)
    ef_good = EventFilterClassifier(features=variables_new_minpt,
                                    dataset_name='hlt1_2{random}',
                                    iterations=5000, sync=False)
    ef_good.fit(good_events, good_events['signal'])
    estimators['top-{} forest preselection'.format(rank)] = ef_good


    

    




KFold prediction using folds column
236727
KFold prediction using folds column
352627

In [33]:

    

forest_base_partial = SklearnClassifier(RandomForestClassifier(n_estimators=300, min_samples_leaf=500, max_depth=7,
                                                               max_features=5))
forest_folding_top_minpt = FoldingClassifier(base_estimator=forest_base_partial, random_state=11, 
                                             features=variables_new, ipc_profile='ssh-ipy')
forest_folding_top_minpt.fit(train, train['signal'])


    

    
Out[33]:





FoldingClassifier(base_estimator=SklearnClassifier(clf=RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
            max_depth=7, max_features=5, max_leaf_nodes=None,
            min_samples_leaf=500, min_samples_split=2,
            min_weight_fraction_leaf=0.0, n_estimators=300, n_jobs=1,
            oob_score=False, random_state=None, verbose=0,
            warm_start=False),
         features=None),
         features=['chi2', 'minpt', 'sumpt', 'fdchi2', 'nlt16'],
         ipc_profile='ssh-ipy', n_folds=2, random_state=11)

In [51]:

    

for rank in range(1, 3):
    good_events = get_best_svr(train, forest_folding_top_minpt, rank)
    ef_good = EventFilterClassifier(features=variables_new,
                                    dataset_name='hlt1_2{random}',
                                    iterations=5000, sync=False)
    ef_good.fit(good_events, good_events['signal'])
    estimators['top-{} forest preselection, minpt'.format(rank)] = ef_good


    

    




KFold prediction using folds column
352627

In [54]:

    

estimators['top-1 forest preselection, minpt'].get_feature_importances()


    

    
Out[54]:






  

    

      

      
effect
      
efficiency
      
information
    
  
  

    

      
chi2
      
 0.449232
      
 0.104
      
 1.000000
    
    

      
minpt
      
 0.485420
      
 0.117
      
 0.955380
    
    

      
sumpt
      
 1.000000
      
 0.247
      
 0.936103
    
    

      
fdchi2
      
 0.769206
      
 0.194
      
 0.918553
    
    

      
nlt16
      
 0.565152
      
 1.000
      
 0.130602
    
  

In [44]:

    

for rank in range(1, 3):
    good_events = get_best_svr(train, forest_folding_top_minpt, rank)
    ef_good = EventFilterClassifier(features=variables_base,
                                    dataset_name='hlt1_2{random}',
                                    iterations=5000, sync=False)
    ef_good.fit(good_events, good_events['signal'])
    estimators['top-{} forest preselection, minpt, base'.format(rank)] = ef_good


    

    




KFold prediction using folds column
352627

In [55]:

    

estimators['top-1 forest preselection, minpt, base'].get_feature_importances()


    

    
Out[55]:






  

    

      

      
effect
      
efficiency
      
information
    
  
  

    

      
mcor
      
 1.000000
      
 0.314
      
 0.920954
    
    

      
chi2
      
 0.273574
      
 0.079
      
 1.000000
    
    

      
pt
      
 0.472722
      
 0.145
      
 0.941393
    
    

      
fdchi2
      
 0.538882
      
 0.170
      
 0.918553
    
    

      
minpt
      
 0.442074
      
 0.134
      
 0.955380
    
    

      
nlt16
      
 0.451962
      
 1.000
      
 0.130602
    
  

In [36]:

    

forest_base_partial = SklearnClassifier(RandomForestClassifier(n_estimators=300, min_samples_leaf=500, max_depth=7,
                                                               max_features=5))
forest_folding_top_mcor = FoldingClassifier(base_estimator=forest_base_partial, random_state=11, 
                                             features=variables_base, ipc_profile='ssh-ipy')
forest_folding_top_mcor.fit(train, train['signal'])


    

    
Out[36]:





FoldingClassifier(base_estimator=SklearnClassifier(clf=RandomForestClassifier(bootstrap=True, class_weight=None, criterion='gini',
            max_depth=7, max_features=5, max_leaf_nodes=None,
            min_samples_leaf=500, min_samples_split=2,
            min_weight_fraction_leaf=0.0, n_estimators=300, n_jobs=1,
            oob_score=False, random_state=None, verbose=0,
            warm_start=False),
         features=None),
         features=['mcor', 'chi2', 'pt', 'fdchi2', 'minpt', 'nlt16'],
         ipc_profile='ssh-ipy', n_folds=2, random_state=11)

In [37]:

    

for rank in range(1, 3):
    good_events = get_best_svr(train, forest_folding_top_mcor, rank)
    ef_good = EventFilterClassifier(features=variables_new,
                                    dataset_name='hlt1_2{random}',
                                    iterations=5000, sync=False)
    ef_good.fit(good_events, good_events['signal'])
    estimators['top-{} forest preselection, minpt; mcor for forest'.format(rank)] = ef_good


    

    




KFold prediction using folds column
236727
KFold prediction using folds column
352627

In [38]:

    

for rank in range(1, 3):
    good_events = get_best_svr(train, forest_folding_top_mcor, rank)
    ef_good = EventFilterClassifier(features=variables_base,
                                    dataset_name='hlt1_2{random}',
                                    iterations=5000, sync=False)
    ef_good.fit(good_events, good_events['signal'])
    estimators['top-{} forest preselection, mcor'.format(rank)] = ef_good


    

    




KFold prediction using folds column
236727
KFold prediction using folds column
352627

In [57]:

    

import cPickle
with open('models/hlt1_body2.pkl', 'w') as file_mn:
    cPickle.dump(estimators, file_mn)


    

In [27]:

    

estimators.keys()


    

    
Out[27]:





['MN BBDT: pt->sumpt, remove minpt',
 'top-1 forest preselection, minpt; mcor for forest',
 'top-1 forest preselection, minpt',
 'MN: pt->sumpt',
 'top-1 forest preselection, minpt, base',
 'top-2 forest preselection, minpt; mcor for forest',
 'additional features',
 'top-1 forest preselection, mcor',
 'top-2 forest preselection, minpt, base',
 'MN: pt->sumpt, remove minpt',
 'top-2 forest preselection, use mcor',
 'MN: mcor cut, mcor var',
 'top-2 forest preselection, minpt',
 'top-2 forest preselection, mcor',
 'MN',
 'MN BBDT: pt->sumpt',
 'top-2 forest preselection',
 'top-1 forest preselection, use mcor',
 'top-1 forest preselection']

In [28]:

    

thresholds = dict()
RATE = [50000.]
events_pass = dict()

for name, cl in estimators.items():
    prob = cl.predict_proba(ds_test_bck)
    if 'mcor cut' not in name and 'additional' not in name:
        thr, result = calculate_thresholds(ds_test_bck, prob, total_test_bck_events, rates=RATE)
        for rate, val in result.items():
            events_pass['{}-{}'.format(rate, name)] = val[1]
    else:
        thr, result = calculate_thresholds(ds_test_bck[ds_test_bck['mcor'] <= 10e3], 
                                           prob[numpy.array(ds_test_bck['mcor']) <= 10e3], 
                                           total_test_bck_events, rates=RATE)
        for rate, val in result.items():
            events_pass['{}-{}'.format(rate, name)] = val[1]

    thresholds[name] = thr
    print name, result


    

    




MN BBDT: pt->sumpt, remove minpt {50000.0: (0.93403029768833201, 2781, 0.04997035200258746)}
top-1 forest preselection, minpt; mcor for forest {50000.0: (0.70734054585716377, 2782, 0.049988320485867786)}
top-1 forest preselection, minpt {50000.0: (0.69145662551796705, 2782, 0.049988320485867786)}
MN: pt->sumpt {50000.0: (0.92164056154030227, 2782, 0.049988320485867786)}
top-1 forest preselection, minpt, base {50000.0: (0.63113157716925705, 2782, 0.049988320485867786)}
top-2 forest preselection, minpt; mcor for forest {50000.0: (0.82364361731341229, 2782, 0.049988320485867786)}
additional features {50000.0: (0.85174856117558884, 2782, 0.049988320485867786)}
top-1 forest preselection, mcor {50000.0: (0.62273222647031767, 2782, 0.049988320485867786)}
top-2 forest preselection, minpt, base {50000.0: (0.77124833132243653, 2782, 0.049988320485867786)}
MN: pt->sumpt, remove minpt {50000.0: (0.92402185981425233, 2782, 0.049988320485867786)}
top-2 forest preselection, use mcor {50000.0: (0.76996316563606948, 2782, 0.049988320485867786)}
MN: mcor cut, mcor var {50000.0: (0.88272822063177037, 2782, 0.049988320485867786)}
top-2 forest preselection, minpt {50000.0: (0.81788106584760678, 2782, 0.049988320485867786)}
top-2 forest preselection, mcor {50000.0: (0.76333098729525073, 2782, 0.049988320485867786)}
MN {50000.0: (0.92181915444909279, 2782, 0.049988320485867786)}
MN BBDT: pt->sumpt {50000.0: (0.93201294998767104, 2770, 0.04977269868650387)}
top-2 forest preselection {50000.0: (0.82176077361122679, 2782, 0.049988320485867786)}
top-1 forest preselection, use mcor {50000.0: (0.63252156232207368, 2782, 0.049988320485867786)}
top-1 forest preselection {50000.0: (0.70730073391624049, 2782, 0.049988320485867786)}

In [35]:

    

est_cut = dict([('MN: mcor cut, mcor var', estimators['MN: mcor cut, mcor var']),
                ('additional features', estimators['additional features'])])
train_modes_eff, statistic = result_statistic(est_cut, sig_train_modes_names, 
                                              ds_test_signal[ds_test_signal['mcor'] <= 10e3],
                                              thresholds, RATE, total_test_signal_events)


    

In [36]:

    

est = []
for key in ['MN', 'MN: pt->sumpt', 'MN: pt->sumpt, remove minpt']:
    est.append((key, estimators[key]))
est = dict(est)

train_modes_eff_rest, statistic_rest = result_statistic(est, sig_train_modes_names, 
                                                        ds_test_signal,
                                                        thresholds, RATE, total_test_signal_events)


    

In [33]:

    

from rep.plotting import BarComparePlot
BarComparePlot(OrderedDict(train_modes_eff.items() + train_modes_eff_rest.items()), 
               sortby=('MN', 50000.0)).plot(new_plot=True, figsize=(24, 8), ylabel='efficiency', fontsize=22)

lgd = legend(bbox_to_anchor=(0.5, 1.4), loc='upper center', ncol=2, fontsize=22)
# plt.savefig('hlt1.pdf' , format='pdf', bbox_extra_artists=(lgd,), bbox_inches='tight')


    

In [37]:

    

compare_hierarhical = OrderedDict()
hierarhical = ['top-1 forest preselection',
               'top-2 forest preselection',
               'top-1 forest preselection, use mcor',
               'top-2 forest preselection, use mcor',


               'top-2 forest preselection, minpt',
               'top-1 forest preselection, minpt',
               'top-2 forest preselection, minpt, base',
               'top-1 forest preselection, minpt, base',
               
               'top-1 forest preselection, minpt; mcor for forest',
               'top-2 forest preselection, minpt; mcor for forest',
               'top-1 forest preselection, mcor',
               'top-2 forest preselection, mcor',
               
               'MN: pt->sumpt',
               'MN BBDT: pt->sumpt', 
               'MN: pt->sumpt, remove minpt',
               'MN BBDT: pt->sumpt, remove minpt'
                ]
for key in hierarhical:
    compare_hierarhical[key] = estimators[key]


train_modes_eff_forest, statistic_forest = result_statistic(compare_hierarhical, sig_train_modes_names, 
                                                            ds_test_signal,
                                                            thresholds, RATE, total_test_signal_events)


    

In [35]:

    

BarComparePlot(OrderedDict([((key, 50000.), train_modes_eff_forest[(key, 50000.)]) for key in hierarhical]),
               sortby=('MN BBDT: pt->sumpt', 50000.0)).plot(new_plot=True, figsize=(24, 8), 
                                                            ylabel='efficiency', fontsize=22)
lgd = legend(bbox_to_anchor=(0.5, 1.7), loc='upper center', ncol=2, fontsize=22)


    

In [36]:

    

pandas.DataFrame(statistic)


    

    
Out[36]:






  

    

      

      
50000.0
      
classifier
      
mode
      
name
    
  
  

    

      
0 
      
 88.580179
      
    additional features
      
 13144001
      
                                                     $B_{s}^{0} \rightarrow \psi(1S)[\mu^{+}\mu^{-}]\phi[K^{+}K^{-}]$
    
    

      
1 
      
 35.416667
      
    additional features
      
 27163400
      
                                                         $D^{*+}(2010) \rightarrow D^{0}[K^{-}\pi^{+}\pi^{0}]\pi^{+}$
    
    

      
2 
      
 86.110714
      
    additional features
      
 11536011
      
 $B^{0} \rightarrow D^{*-}[\pi^{-}\bar{D}^{0}[K^{+}\pi^{-}]\tau^{+}[\pi^{+}\pi^{+}\pi^{-}\bar{\nu}_{\tau}]\nu_{\tau}$
    
    

      
3 
      
 92.829343
      
    additional features
      
 11296013
      
                                             $B^{0} \rightarrow D^{+}[K^{-}\pi^{+}\pi^{+}]D^{-}[K^{+}\pi^{-}\pi^{-}]$
    
    

      
4 
      
 86.953969
      
    additional features
      
 11114001
      
                                                                $B^{0} \rightarrow K^{*}[K^{+}\pi^{-}]\mu^{+}\mu^{-}$
    
    

      
5 
      
 87.452168
      
    additional features
      
 13264021
      
                                                          $B_{s}^{0} \rightarrow D_{s}^{-}[K^{+}K^{-}\pi^{-}]\pi^{+}$
    
    

      
6 
      
 63.327370
      
    additional features
      
 21263002
      
                                                                                $D^{+} \rightarrow K^{-}K^{+}\pi^{+}$
    
    

      
7 
      
 77.249895
      
    additional features
      
 12873002
      
                                                        $B^{+} \rightarrow \bar{D}^{0}[K^{+}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
8 
      
 90.046757
      
    additional features
      
 13246001
      
                                             $B_{s}^{0} \rightarrow \psi(1S)[\mu^{+}\mu^{-}]K^{+}K^{-}\pi^{+}\pi^{-}$
    
    

      
9 
      
 81.376582
      
    additional features
      
 13774002
      
                                                 $B_{s}^{0} \rightarrow D_{s}^{-}[K^{+}K^{-}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
10
      
 76.051925
      
    additional features
      
 11102003
      
                                                                                     $B^{0} \rightarrow K^{-}\pi^{+}$
    
    

      
11
      
 82.209155
      
    additional features
      
 13112001
      
                                                                               $B_{s}^{0} \rightarrow \mu^{+}\mu^{-}$
    
    

      
12
      
 64.442603
      
    additional features
      
 13512010
      
                                                                        $B_{s}^{0} \rightarrow K^{-}\mu^{+}\nu_{\mu}$
    
    

      
13
      
 61.794983
      
    additional features
      
 15512011
      
                                                            $\Lambda_{b}^{0} \rightarrow p^{+}\mu^{-}\bar{\nu}_{\mu}$
    
    

      
14
      
 57.724393
      
    additional features
      
 13102412
      
                                                                            $B_{s}^{0} \rightarrow K^{-}K^{+}\pi^{0}$
    
    

      
15
      
 25.659691
      
    additional features
      
 27265101
      
                                             $D^{*+}(2010) \rightarrow D^{0}[K_{S}[\pi^{+}\pi^{-}]K^{+}K^{-}]\pi^{+}$
    
    

      
16
      
  8.943383
      
    additional features
      
 12103121
      
                                                                       $B^{+} \rightarrow K^{+}K_{S}[\pi^{+}\pi^{-}]$
    
    

      
17
      
 87.179487
      
    additional features
      
 13144020
      
                                                     $B_{s}^{0} \rightarrow \psi(2S)[\mu^{+}\mu^{-}]\phi[K^{+}K^{-}]$
    
    

      
18
      
 52.680653
      
    additional features
      
 27163002
      
                                                                  $D^{*+}(2010) \rightarrow D^{0}[K^{+}K^{-}]\pi^{+}$
    
    

      
19
      
 88.460456
      
    additional features
      
 15164001
      
                                              $\Lambda_{b}^{0} \rightarrow \Lambda_{c}^{+}[p^{+}K^{+}\pi^{-}]\pi^{-}$
    
    

      
20
      
 43.584071
      
    additional features
      
 27265001
      
                                                    $D^{*+}(2010) \rightarrow D^{0}[K^{+}K^{-}\pi^{+}\pi^{-}]\pi^{+}$
    
    

      
21
      
 81.857819
      
    additional features
      
 11874004
      
                                         $B^{0} \rightarrow D^{*-}[\pi^{-}\bar{D}^{0}[K^{+}\pi^{-}]]\mu^{+}\nu_{\mu}$
    
    

      
22
      
 66.620924
      
    additional features
      
 11114101
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\mu^{+}\mu^{-}$
    
    

      
23
      
 62.097583
      
    additional features
      
 11104121
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\pi^{+}\pi^{-}$
    
    

      
24
      
 79.610195
      
    additional features
      
 11874042
      
                                                       $B^{0} \rightarrow D^{-}[K^{+}\pi^{-}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
25
      
 84.465021
      
    additional features
      
 12103035
      
                                                                                $B^{+} \rightarrow \pi^{+}K^{-}K^{+}$
    
    

      
26
      
 88.155856
      
 MN: mcor cut, mcor var
      
 13144001
      
                                                     $B_{s}^{0} \rightarrow \psi(1S)[\mu^{+}\mu^{-}]\phi[K^{+}K^{-}]$
    
    

      
27
      
 33.854167
      
 MN: mcor cut, mcor var
      
 27163400
      
                                                         $D^{*+}(2010) \rightarrow D^{0}[K^{-}\pi^{+}\pi^{0}]\pi^{+}$
    
    

      
28
      
 85.910456
      
 MN: mcor cut, mcor var
      
 11536011
      
 $B^{0} \rightarrow D^{*-}[\pi^{-}\bar{D}^{0}[K^{+}\pi^{-}]\tau^{+}[\pi^{+}\pi^{+}\pi^{-}\bar{\nu}_{\tau}]\nu_{\tau}$
    
    

      
29
      
 92.605260
      
 MN: mcor cut, mcor var
      
 11296013
      
                                             $B^{0} \rightarrow D^{+}[K^{-}\pi^{+}\pi^{+}]D^{-}[K^{+}\pi^{-}\pi^{-}]$
    
    

      
30
      
 86.686458
      
 MN: mcor cut, mcor var
      
 11114001
      
                                                                $B^{0} \rightarrow K^{*}[K^{+}\pi^{-}]\mu^{+}\mu^{-}$
    
    

      
31
      
 87.396365
      
 MN: mcor cut, mcor var
      
 13264021
      
                                                          $B_{s}^{0} \rightarrow D_{s}^{-}[K^{+}K^{-}\pi^{-}]\pi^{+}$
    
    

      
32
      
 63.550984
      
 MN: mcor cut, mcor var
      
 21263002
      
                                                                                $D^{+} \rightarrow K^{-}K^{+}\pi^{+}$
    
    

      
33
      
 76.956886
      
 MN: mcor cut, mcor var
      
 12873002
      
                                                        $B^{+} \rightarrow \bar{D}^{0}[K^{+}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
34
      
 89.772666
      
 MN: mcor cut, mcor var
      
 13246001
      
                                             $B_{s}^{0} \rightarrow \psi(1S)[\mu^{+}\mu^{-}]K^{+}K^{-}\pi^{+}\pi^{-}$
    
    

      
35
      
 81.186709
      
 MN: mcor cut, mcor var
      
 13774002
      
                                                 $B_{s}^{0} \rightarrow D_{s}^{-}[K^{+}K^{-}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
36
      
 75.881826
      
 MN: mcor cut, mcor var
      
 11102003
      
                                                                                     $B^{0} \rightarrow K^{-}\pi^{+}$
    
    

      
37
      
 82.013144
      
 MN: mcor cut, mcor var
      
 13112001
      
                                                                               $B_{s}^{0} \rightarrow \mu^{+}\mu^{-}$
    
    

      
38
      
 64.350509
      
 MN: mcor cut, mcor var
      
 13512010
      
                                                                        $B_{s}^{0} \rightarrow K^{-}\mu^{+}\nu_{\mu}$
    
    

      
39
      
 61.700739
      
 MN: mcor cut, mcor var
      
 15512011
      
                                                            $\Lambda_{b}^{0} \rightarrow p^{+}\mu^{-}\bar{\nu}_{\mu}$
    
    

      
40
      
 57.624036
      
 MN: mcor cut, mcor var
      
 13102412
      
                                                                            $B_{s}^{0} \rightarrow K^{-}K^{+}\pi^{0}$
    
    

      
41
      
 24.385805
      
 MN: mcor cut, mcor var
      
 27265101
      
                                             $D^{*+}(2010) \rightarrow D^{0}[K_{S}[\pi^{+}\pi^{-}]K^{+}K^{-}]\pi^{+}$
    
    

      
42
      
  8.600793
      
 MN: mcor cut, mcor var
      
 12103121
      
                                                                       $B^{+} \rightarrow K^{+}K_{S}[\pi^{+}\pi^{-}]$
    
    

      
43
      
 86.884412
      
 MN: mcor cut, mcor var
      
 13144020
      
                                                     $B_{s}^{0} \rightarrow \psi(2S)[\mu^{+}\mu^{-}]\phi[K^{+}K^{-}]$
    
    

      
44
      
 52.447552
      
 MN: mcor cut, mcor var
      
 27163002
      
                                                                  $D^{*+}(2010) \rightarrow D^{0}[K^{+}K^{-}]\pi^{+}$
    
    

      
45
      
 88.404165
      
 MN: mcor cut, mcor var
      
 15164001
      
                                              $\Lambda_{b}^{0} \rightarrow \Lambda_{c}^{+}[p^{+}K^{+}\pi^{-}]\pi^{-}$
    
    

      
46
      
 42.699115
      
 MN: mcor cut, mcor var
      
 27265001
      
                                                    $D^{*+}(2010) \rightarrow D^{0}[K^{+}K^{-}\pi^{+}\pi^{-}]\pi^{+}$
    
    

      
47
      
 81.695241
      
 MN: mcor cut, mcor var
      
 11874004
      
                                         $B^{0} \rightarrow D^{*-}[\pi^{-}\bar{D}^{0}[K^{+}\pi^{-}]]\mu^{+}\nu_{\mu}$
    
    

      
48
      
 66.379405
      
 MN: mcor cut, mcor var
      
 11114101
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\mu^{+}\mu^{-}$
    
    

      
49
      
 61.869585
      
 MN: mcor cut, mcor var
      
 11104121
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\pi^{+}\pi^{-}$
    
    

      
50
      
 79.460270
      
 MN: mcor cut, mcor var
      
 11874042
      
                                                       $B^{0} \rightarrow D^{-}[K^{+}\pi^{-}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
51
      
 84.370054
      
 MN: mcor cut, mcor var
      
 12103035
      
                                                                                $B^{+} \rightarrow \pi^{+}K^{-}K^{+}$
    
  

In [37]:

    

pandas.DataFrame(statistic_rest)


    

    
Out[37]:






  

    

      

      
50000.0
      
classifier
      
mode
      
name
    
  
  

    

      
0 
      
 87.528596
      
                          MN
      
 13144001
      
                                                     $B_{s}^{0} \rightarrow \psi(1S)[\mu^{+}\mu^{-}]\phi[K^{+}K^{-}]$
    
    

      
1 
      
 31.770833
      
                          MN
      
 27163400
      
                                                         $D^{*+}(2010) \rightarrow D^{0}[K^{-}\pi^{+}\pi^{0}]\pi^{+}$
    
    

      
2 
      
 85.037906
      
                          MN
      
 11536011
      
 $B^{0} \rightarrow D^{*-}[\pi^{-}\bar{D}^{0}[K^{+}\pi^{-}]\tau^{+}[\pi^{+}\pi^{+}\pi^{-}\bar{\nu}_{\tau}]\nu_{\tau}$
    
    

      
3 
      
 92.251445
      
                          MN
      
 11296013
      
                                             $B^{0} \rightarrow D^{+}[K^{-}\pi^{+}\pi^{+}]D^{-}[K^{+}\pi^{-}\pi^{-}]$
    
    

      
4 
      
 86.190585
      
                          MN
      
 11114001
      
                                                                $B^{0} \rightarrow K^{*}[K^{+}\pi^{-}]\mu^{+}\mu^{-}$
    
    

      
5 
      
 87.244898
      
                          MN
      
 13264021
      
                                                          $B_{s}^{0} \rightarrow D_{s}^{-}[K^{+}K^{-}\pi^{-}]\pi^{+}$
    
    

      
6 
      
 61.694991
      
                          MN
      
 21263002
      
                                                                                $D^{+} \rightarrow K^{-}K^{+}\pi^{+}$
    
    

      
7 
      
 77.040603
      
                          MN
      
 12873002
      
                                                        $B^{+} \rightarrow \bar{D}^{0}[K^{+}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
8 
      
 88.789165
      
                          MN
      
 13246001
      
                                             $B_{s}^{0} \rightarrow \psi(1S)[\mu^{+}\mu^{-}]K^{+}K^{-}\pi^{+}\pi^{-}$
    
    

      
9 
      
 80.411392
      
                          MN
      
 13774002
      
                                                 $B_{s}^{0} \rightarrow D_{s}^{-}[K^{+}K^{-}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
10
      
 76.741271
      
                          MN
      
 11102003
      
                                                                                     $B^{0} \rightarrow K^{-}\pi^{+}$
    
    

      
11
      
 82.589646
      
                          MN
      
 13112001
      
                                                                               $B_{s}^{0} \rightarrow \mu^{+}\mu^{-}$
    
    

      
12
      
 64.355114
      
                          MN
      
 13512010
      
                                                                        $B_{s}^{0} \rightarrow K^{-}\mu^{+}\nu_{\mu}$
    
    

      
13
      
 61.932724
      
                          MN
      
 15512011
      
                                                            $\Lambda_{b}^{0} \rightarrow p^{+}\mu^{-}\bar{\nu}_{\mu}$
    
    

      
14
      
 58.094462
      
                          MN
      
 13102412
      
                                                                            $B_{s}^{0} \rightarrow K^{-}K^{+}\pi^{0}$
    
    

      
15
      
 22.929936
      
                          MN
      
 27265101
      
                                             $D^{*+}(2010) \rightarrow D^{0}[K_{S}[\pi^{+}\pi^{-}]K^{+}K^{-}]\pi^{+}$
    
    

      
16
      
  9.123693
      
                          MN
      
 12103121
      
                                                                       $B^{+} \rightarrow K^{+}K_{S}[\pi^{+}\pi^{-}]$
    
    

      
17
      
 86.284086
      
                          MN
      
 13144020
      
                                                     $B_{s}^{0} \rightarrow \psi(2S)[\mu^{+}\mu^{-}]\phi[K^{+}K^{-}]$
    
    

      
18
      
 50.582751
      
                          MN
      
 27163002
      
                                                                  $D^{*+}(2010) \rightarrow D^{0}[K^{+}K^{-}]\pi^{+}$
    
    

      
19
      
 87.981987
      
                          MN
      
 15164001
      
                                              $\Lambda_{b}^{0} \rightarrow \Lambda_{c}^{+}[p^{+}K^{+}\pi^{-}]\pi^{-}$
    
    

      
20
      
 39.823009
      
                          MN
      
 27265001
      
                                                    $D^{*+}(2010) \rightarrow D^{0}[K^{+}K^{-}\pi^{+}\pi^{-}]\pi^{+}$
    
    

      
21
      
 81.510494
      
                          MN
      
 11874004
      
                                         $B^{0} \rightarrow D^{*-}[\pi^{-}\bar{D}^{0}[K^{+}\pi^{-}]]\mu^{+}\nu_{\mu}$
    
    

      
22
      
 66.412339
      
                          MN
      
 11114101
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\mu^{+}\mu^{-}$
    
    

      
23
      
 62.088463
      
                          MN
      
 11104121
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\pi^{+}\pi^{-}$
    
    

      
24
      
 79.260370
      
                          MN
      
 11874042
      
                                                       $B^{0} \rightarrow D^{-}[K^{+}\pi^{-}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
25
      
 84.338398
      
                          MN
      
 12103035
      
                                                                                $B^{+} \rightarrow \pi^{+}K^{-}K^{+}$
    
    

      
26
      
 87.524906
      
 MN: pt->sumpt, remove minpt
      
 13144001
      
                                                     $B_{s}^{0} \rightarrow \psi(1S)[\mu^{+}\mu^{-}]\phi[K^{+}K^{-}]$
    
    

      
27
      
 31.770833
      
 MN: pt->sumpt, remove minpt
      
 27163400
      
                                                         $D^{*+}(2010) \rightarrow D^{0}[K^{-}\pi^{+}\pi^{0}]\pi^{+}$
    
    

      
28
      
 84.923473
      
 MN: pt->sumpt, remove minpt
      
 11536011
      
 $B^{0} \rightarrow D^{*-}[\pi^{-}\bar{D}^{0}[K^{+}\pi^{-}]\tau^{+}[\pi^{+}\pi^{+}\pi^{-}\bar{\nu}_{\tau}]\nu_{\tau}$
    
    

      
29
      
 92.275032
      
 MN: pt->sumpt, remove minpt
      
 11296013
      
                                             $B^{0} \rightarrow D^{+}[K^{-}\pi^{+}\pi^{+}]D^{-}[K^{+}\pi^{-}\pi^{-}]$
    
    

      
...
      
...
      
...
      
...
      
...
    
    

      
48
      
 66.467230
      
 MN: pt->sumpt, remove minpt
      
 11114101
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\mu^{+}\mu^{-}$
    
    

      
49
      
 62.061104
      
 MN: pt->sumpt, remove minpt
      
 11104121
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\pi^{+}\pi^{-}$
    
    

      
50
      
 79.260370
      
 MN: pt->sumpt, remove minpt
      
 11874042
      
                                                       $B^{0} \rightarrow D^{-}[K^{+}\pi^{-}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
51
      
 84.544160
      
 MN: pt->sumpt, remove minpt
      
 12103035
      
                                                                                $B^{+} \rightarrow \pi^{+}K^{-}K^{+}$
    
    

      
52
      
 87.679876
      
               MN: pt->sumpt
      
 13144001
      
                                                     $B_{s}^{0} \rightarrow \psi(1S)[\mu^{+}\mu^{-}]\phi[K^{+}K^{-}]$
    
    

      
53
      
 31.770833
      
               MN: pt->sumpt
      
 27163400
      
                                                         $D^{*+}(2010) \rightarrow D^{0}[K^{-}\pi^{+}\pi^{0}]\pi^{+}$
    
    

      
54
      
 84.994994
      
               MN: pt->sumpt
      
 11536011
      
 $B^{0} \rightarrow D^{*-}[\pi^{-}\bar{D}^{0}[K^{+}\pi^{-}]\tau^{+}[\pi^{+}\pi^{+}\pi^{-}\bar{\nu}_{\tau}]\nu_{\tau}$
    
    

      
55
      
 92.345795
      
               MN: pt->sumpt
      
 11296013
      
                                             $B^{0} \rightarrow D^{+}[K^{-}\pi^{+}\pi^{+}]D^{-}[K^{+}\pi^{-}\pi^{-}]$
    
    

      
56
      
 86.396111
      
               MN: pt->sumpt
      
 11114001
      
                                                                $B^{0} \rightarrow K^{*}[K^{+}\pi^{-}]\mu^{+}\mu^{-}$
    
    

      
57
      
 87.149235
      
               MN: pt->sumpt
      
 13264021
      
                                                          $B_{s}^{0} \rightarrow D_{s}^{-}[K^{+}K^{-}\pi^{-}]\pi^{+}$
    
    

      
58
      
 61.314848
      
               MN: pt->sumpt
      
 21263002
      
                                                                                $D^{+} \rightarrow K^{-}K^{+}\pi^{+}$
    
    

      
59
      
 77.103391
      
               MN: pt->sumpt
      
 12873002
      
                                                        $B^{+} \rightarrow \bar{D}^{0}[K^{+}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
60
      
 88.837534
      
               MN: pt->sumpt
      
 13246001
      
                                             $B_{s}^{0} \rightarrow \psi(1S)[\mu^{+}\mu^{-}]K^{+}K^{-}\pi^{+}\pi^{-}$
    
    

      
61
      
 80.474684
      
               MN: pt->sumpt
      
 13774002
      
                                                 $B_{s}^{0} \rightarrow D_{s}^{-}[K^{+}K^{-}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
62
      
 76.839749
      
               MN: pt->sumpt
      
 11102003
      
                                                                                     $B^{0} \rightarrow K^{-}\pi^{+}$
    
    

      
63
      
 82.860602
      
               MN: pt->sumpt
      
 13112001
      
                                                                               $B_{s}^{0} \rightarrow \mu^{+}\mu^{-}$
    
    

      
64
      
 64.511673
      
               MN: pt->sumpt
      
 13512010
      
                                                                        $B_{s}^{0} \rightarrow K^{-}\mu^{+}\nu_{\mu}$
    
    

      
65
      
 62.106713
      
               MN: pt->sumpt
      
 15512011
      
                                                            $\Lambda_{b}^{0} \rightarrow p^{+}\mu^{-}\bar{\nu}_{\mu}$
    
    

      
66
      
 58.012921
      
               MN: pt->sumpt
      
 13102412
      
                                                                            $B_{s}^{0} \rightarrow K^{-}K^{+}\pi^{0}$
    
    

      
67
      
 22.838944
      
               MN: pt->sumpt
      
 27265101
      
                                             $D^{*+}(2010) \rightarrow D^{0}[K_{S}[\pi^{+}\pi^{-}]K^{+}K^{-}]\pi^{+}$
    
    

      
68
      
  9.087631
      
               MN: pt->sumpt
      
 12103121
      
                                                                       $B^{+} \rightarrow K^{+}K_{S}[\pi^{+}\pi^{-}]$
    
    

      
69
      
 86.523199
      
               MN: pt->sumpt
      
 13144020
      
                                                     $B_{s}^{0} \rightarrow \psi(2S)[\mu^{+}\mu^{-}]\phi[K^{+}K^{-}]$
    
    

      
70
      
 50.349650
      
               MN: pt->sumpt
      
 27163002
      
                                                                  $D^{*+}(2010) \rightarrow D^{0}[K^{+}K^{-}]\pi^{+}$
    
    

      
71
      
 87.869406
      
               MN: pt->sumpt
      
 15164001
      
                                              $\Lambda_{b}^{0} \rightarrow \Lambda_{c}^{+}[p^{+}K^{+}\pi^{-}]\pi^{-}$
    
    

      
72
      
 38.716814
      
               MN: pt->sumpt
      
 27265001
      
                                                    $D^{*+}(2010) \rightarrow D^{0}[K^{+}K^{-}\pi^{+}\pi^{-}]\pi^{+}$
    
    

      
73
      
 81.628732
      
               MN: pt->sumpt
      
 11874004
      
                                         $B^{0} \rightarrow D^{*-}[\pi^{-}\bar{D}^{0}[K^{+}\pi^{-}]]\mu^{+}\nu_{\mu}$
    
    

      
74
      
 66.533099
      
               MN: pt->sumpt
      
 11114101
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\mu^{+}\mu^{-}$
    
    

      
75
      
 62.088463
      
               MN: pt->sumpt
      
 11104121
      
                                                          $B^{0} \rightarrow K_{S}^{0}[\pi^{+}\pi^{-}]\pi^{+}\pi^{-}$
    
    

      
76
      
 79.310345
      
               MN: pt->sumpt
      
 11874042
      
                                                       $B^{0} \rightarrow D^{-}[K^{+}\pi^{-}\pi^{-}]\mu^{+}\nu_{\mu}$
    
    

      
77
      
 84.496676
      
               MN: pt->sumpt
      
 12103035
      
                                                                                $B^{+} \rightarrow \pi^{+}K^{-}K^{+}$
    
  

78 rows × 4 columns

In [38]:

    

from rep.data import LabeledDataStorage
from rep.report import ClassificationReport
lds = LabeledDataStorage(test, test['signal'])
report = ClassificationReport(OrderedDict(est.items() + compare_hierarhical.items()), lds)


    

In [39]:

    

report.roc().plot(new_plot=True)


    

In [40]:

    

mode_metrics = dict()
for mode in sig_train_modes_names:
    mode_metrics[mode] = dict()
    for rate in [50000.]:
        mode_metrics[mode][rate] = generate_topo_metric(ds_test_bck, ds_test_signal[ds_test_signal['mode'] == mode], 
                                                        total_test_bck_events, total_test_signal_events[mode], rate)


    

In [41]:

    

def generate_select_mode(mode):
    def select_mode(pd):
        result = numpy.zeros(len(pd), dtype=bool)
        result[numpy.array(pd['mode']) == mode] = True
        result[numpy.array(pd['signal']) == 0] = True
        return result
    return select_mode


    

In [42]:

    

staged_eff_plots = []
for mode in sig_train_modes_names:
    eff_metrics = OrderedDict()
    select_mode = generate_select_mode(mode)
    if len(ds_test_signal[ds_test_signal['mode'] == mode]) <= 0:
        continue
    staged_eff_plots.append(report.learning_curve(mode_metrics[mode][50000.], mask=select_mode, steps=10,
                                                  metric_label='mode {}, rate {}, eff'.format(mode, 50000.)))


    

In [43]:

    

for n, elem in enumerate(staged_eff_plots):
    elem.plot(new_plot=True)


    

    




/moosefs/ipython_env/local/lib/python2.7/site-packages/matplotlib/pyplot.py:412: RuntimeWarning: More than 20 figures have been opened. Figures created through the pyplot interface (`matplotlib.pyplot.figure`) are retained until explicitly closed and may consume too much memory. (To control this warning, see the rcParam `figure.max_num_figures`).
  max_open_warning, RuntimeWarning)






    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    













    

In [29]:

    

not_passed_event_sig = 0
not_passed_event_bck = (total_bck_events - total_bck_events_presel) // 2

for mode in sig_train_modes_names:
    not_passed_event_sig += (total_signal_events_by_mode[mode] - total_signal_events_by_mode_presel[mode]) // 2


    

In [55]:

    

plots = OrderedDict()
for key, value in est.items() + compare_hierarhical.items():
    plots[key] = plot_roc_events(value, ds_test_signal, ds_test_bck, key, False, not_passed_event_sig,
                                 not_passed_event_bck)


    

    




MN AUC: 0.864113294573
MN: pt->sumpt, remove minpt AUC: 0.864273024603
MN: pt->sumpt AUC: 0.864540490078
top-1 forest preselection AUC: 0.864459518369
top-2 forest preselection AUC: 0.864471418886
top-1 forest preselection, use mcor AUC: 0.867742761199
top-2 forest preselection, use mcor AUC: 0.867691311986
top-2 forest preselection, minpt AUC: 0.86470551744
top-1 forest preselection, minpt AUC: 0.864707707735
top-2 forest preselection, minpt, base AUC: 0.867696534292
top-1 forest preselection, minpt, base AUC: 0.867750281859
top-1 forest preselection, minpt; mcor for forest AUC: 0.864697061289
top-2 forest preselection, minpt; mcor for forest AUC: 0.864681445296
top-1 forest preselection, mcor AUC: 0.867740130257
top-2 forest preselection, mcor AUC: 0.867689542548
MN: pt->sumpt AUC: 0.864540490078
MN BBDT: pt->sumpt AUC: 0.86414890354
MN: pt->sumpt, remove minpt AUC: 0.864273024603
MN BBDT: pt->sumpt, remove minpt AUC: 0.863870455117

In [57]:

    

from rep.plotting import FunctionsPlot
FunctionsPlot(plots).plot(new_plot=True, xlim=(0, 0.1), ylim=(0., 1))
plot([1. * events_pass['50000.0-MN'] / (statistic_length(ds_test_bck)['Events'] + not_passed_event_bck)] * 2, [0., 1], 'b--', label='rate: 50 kHz')
lgd = legend(loc='upper center', fontsize=16, bbox_to_anchor=(0.5, 1.7), ncol=2)


    

In [44]:

    

plots = OrderedDict()
for key, value in est.items() + compare_hierarhical.items():
    plots[key] = plot_roc_events(value, ds_test_signal, ds_test_bck, key, not_passed_event_sig, not_passed_event_bck)


    

    




MN AUC: 0.896901106839
MN: pt->sumpt, remove minpt AUC: 0.895331858542
MN: pt->sumpt AUC: 0.899213091883
top-1 forest preselection AUC: 0.896316326496
top-2 forest preselection AUC: 0.896407672672
top-1 forest preselection, use mcor AUC: 0.91610524636
top-2 forest preselection, use mcor AUC: 0.915107933082
top-2 forest preselection, minpt AUC: 0.899660469928
top-1 forest preselection, minpt AUC: 0.900116758229
top-2 forest preselection, minpt, base AUC: 0.915294058153
top-1 forest preselection, minpt, base AUC: 0.916469969077
top-1 forest preselection, minpt; mcor for forest AUC: 0.900497192316
top-2 forest preselection, minpt; mcor for forest AUC: 0.899572796211
top-1 forest preselection, mcor AUC: 0.912791689117
top-2 forest preselection, mcor AUC: 0.911954421408
MN: pt->sumpt AUC: 0.899213091883
MN BBDT: pt->sumpt AUC: 0.893164462992
MN: pt->sumpt, remove minpt AUC: 0.895331858542
MN BBDT: pt->sumpt, remove minpt AUC: 0.889448537654

In [45]:

    

from rep.plotting import FunctionsPlot
FunctionsPlot(plots).plot(new_plot=True, xlim=(0, 1), ylim=(0., 1))
plot([1. * events_pass['50000.0-MN'] / (statistic_length(ds_test_bck)['Events'] + not_passed_event_bck)] * 2, [0., 1], 'b--', label='rate: 50 kHz')
lgd = legend(loc='upper center', fontsize=16, bbox_to_anchor=(0.5, 1.7), ncol=2)


    

In [51]:

    

FunctionsPlot(plots).plot(new_plot=True, xlim=(0.42, 0.44), ylim=(0.945, 0.965))
plot([1. * events_pass['50000.0-MN'] / statistic_length(ds_test_bck)['Events']] * 2, [0., 1], 'b--', label='rate: 50 kHz')
lgd = legend(loc='upper center', fontsize=16, bbox_to_anchor=(0.5, 1.7), ncol=2)


    

In [46]:

    

plots_unnormed = OrderedDict()
for key, value in est.items() + compare_hierarhical.items():
    plots_unnormed[key] = plot_roc_events(value, ds_test_signal, ds_test_bck, key, normed_channels=False)


    

    




MN AUC: 0.922813122724
MN: pt->sumpt, remove minpt AUC: 0.924596862336
MN: pt->sumpt AUC: 0.927583706847
top-1 forest preselection AUC: 0.926679478354
top-2 forest preselection AUC: 0.926812373982
top-1 forest preselection, use mcor AUC: 0.963344157544
top-2 forest preselection, use mcor AUC: 0.962769613112
top-2 forest preselection, minpt AUC: 0.929426602909
top-1 forest preselection, minpt AUC: 0.929451062407
top-2 forest preselection, minpt, base AUC: 0.962827931726
top-1 forest preselection, minpt, base AUC: 0.963428142372
top-1 forest preselection, minpt; mcor for forest AUC: 0.929332171253
top-2 forest preselection, minpt; mcor for forest AUC: 0.929157784102
top-1 forest preselection, mcor AUC: 0.96331477725
top-2 forest preselection, mcor AUC: 0.962749853422
MN: pt->sumpt AUC: 0.927583706847
MN BBDT: pt->sumpt AUC: 0.923210775718
MN: pt->sumpt, remove minpt AUC: 0.924596862336
MN BBDT: pt->sumpt, remove minpt AUC: 0.920101282256

In [47]:

    

from rep.plotting import FunctionsPlot
FunctionsPlot(plots_unnormed).plot(new_plot=True, xlim=(0, 1), ylim=(0., 1))
plot([1. * events_pass['50000.0-MN'] / statistic_length(ds_test_bck)['Events']] * 2, [0., 1], 'b--', label='rate: 50 kHz')
lgd = legend(loc='upper center', fontsize=16, bbox_to_anchor=(0.5, 1.7), ncol=2)


    

In [55]:

    

with open("bbdt_run2/hlt1_borders_sumpt.mx", "w") as f:
    f.write(estimators['MN BBDT: pt->sumpt'].formula_mx)


    

In [56]:

    

with open("bbdt_run2/hlt1_borders_minpt.mx", "w") as f:
    f.write(estimators['MN BBDT: pt->sumpt, remove minpt'].formula_mx)