In [1]:

    

import matplotlib.pyplot as plt
import numpy as np
import pandas as pd

import composition as comp

%matplotlib inline


    

    




/home/jbourbeau/.local/lib/python2.7/site-packages/sklearn/cross_validation.py:44: DeprecationWarning: This module was deprecated in version 0.18 in favor of the model_selection module into which all the refactored classes and functions are moved. Also note that the interface of the new CV iterators are different from that of this module. This module will be removed in 0.20.
  "This module will be removed in 0.20.", DeprecationWarning)

In [2]:

    

feature_list, feature_labels = comp.get_training_features()


    

In [3]:

    

pipeline_str = 'xgboost'
scoring = 'accuracy'
cv = 10


    

In [4]:

    

sfs = pd.read_csv('SFS-results/{}_forward_nofloat_{}_cv{}.csv'.format(pipeline_str, scoring, cv))
sffs = pd.read_csv('SFS-results/{}_forward_floating_{}_cv{}.csv'.format(pipeline_str, scoring, cv))
sbs = pd.read_csv('SFS-results/{}_backward_nofloat_{}_cv{}.csv'.format(pipeline_str, scoring, cv))
sfbs = pd.read_csv('SFS-results/{}_backward_floating_{}_cv{}.csv'.format(pipeline_str, scoring, cv))


    

In [5]:

    

sfs


    

    
Out[5]:






  

    

      

      
Unnamed: 0
      
avg_score
      
ci_bound
      
cv_scores
      
feature_idx
      
std_dev
      
std_err
    
  
  

    

      
0
      
1
      
0.615184
      
0.002381
      
[ 0.61324942  0.62146406  0.61371315  0.609738...
      
(12,)
      
0.003206
      
0.001069
    
    

      
1
      
2
      
0.770208
      
0.002141
      
[ 0.76442531  0.76787016  0.77250745  0.773964...
      
(9, 12)
      
0.002882
      
0.000961
    
    

      
2
      
3
      
0.777038
      
0.002104
      
[ 0.77310368  0.7736999   0.77793972  0.782378...
      
(9, 12, 5)
      
0.002833
      
0.000944
    
    

      
3
      
4
      
0.780642
      
0.002175
      
[ 0.77575356  0.77913216  0.78032461  0.783040...
      
(9, 12, 5, 4)
      
0.002928
      
0.000976
    
    

      
4
      
5
      
0.782477
      
0.002080
      
[ 0.77827095  0.7783372   0.78403445  0.785226...
      
(9, 2, 12, 5, 4)
      
0.002800
      
0.000933
    
    

      
5
      
6
      
0.783491
      
0.002046
      
[ 0.77873468  0.77972839  0.78363697  0.786684...
      
(1, 2, 4, 5, 9, 12)
      
0.002754
      
0.000918
    
    

      
6
      
7
      
0.784206
      
0.002183
      
[ 0.77899967  0.77999338  0.78482941  0.787810...
      
(1, 2, 4, 5, 9, 12, 14)
      
0.002940
      
0.000980
    
    

      
7
      
8
      
0.784511
      
0.001996
      
[ 0.77966214  0.78078834  0.78376946  0.787810...
      
(1, 2, 4, 5, 9, 11, 12, 14)
      
0.002687
      
0.000896
    
    

      
8
      
9
      
0.784564
      
0.002013
      
[ 0.77966214  0.78025836  0.78575687  0.788075...
      
(0, 1, 2, 4, 5, 9, 11, 12, 14)
      
0.002710
      
0.000903
    
    

      
9
      
10
      
0.784564
      
0.002028
      
[ 0.77959589  0.78025836  0.78575687  0.788141...
      
(0, 1, 2, 4, 5, 6, 9, 11, 12, 14)
      
0.002731
      
0.000910
    
    

      
10
      
11
      
0.784564
      
0.002028
      
[ 0.77959589  0.78025836  0.78575687  0.788141...
      
(0, 1, 2, 4, 5, 6, 9, 11, 12, 13, 14)
      
0.002731
      
0.000910
    
    

      
11
      
12
      
0.784564
      
0.002028
      
[ 0.77959589  0.78025836  0.78575687  0.788141...
      
(0, 1, 2, 4, 5, 6, 9, 11, 12, 13, 14, 15)
      
0.002731
      
0.000910
    
    

      
12
      
13
      
0.784233
      
0.002003
      
[ 0.7793309   0.78045711  0.78476317  0.788141...
      
(0, 1, 2, 4, 5, 6, 7, 9, 11, 12, 13, 14, 15)
      
0.002696
      
0.000899
    
    

      
13
      
14
      
0.784716
      
0.002262
      
[ 0.78019212  0.78138456  0.7852269   0.788605...
      
(0, 1, 2, 4, 5, 6, 7, 8, 9, 11, 12, 13, 14, 15)
      
0.003046
      
0.001015
    
    

      
14
      
15
      
0.784756
      
0.002156
      
[ 0.77986088  0.78191454  0.78535939  0.789069...
      
(0, 1, 2, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14...
      
0.002902
      
0.000967
    
    

      
15
      
16
      
0.784756
      
0.002156
      
[ 0.77986088  0.78191454  0.78535939  0.789069...
      
(0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13,...
      
0.002902
      
0.000967
    
  

In [6]:

    

selector_list = [sfs, sffs, sbs, sfbs]
selector_label = ['SFS', 'SFFS', 'SBS', 'SFBS']
fig, ax = plt.subplots()
for idx, selector in enumerate(selector_list):
    scores = [row['avg_score'] for index, row in selector.iterrows()]
    errs = [row['std_dev'] for index, row in selector.iterrows()]
    ax.errorbar(range(1, len(selector)+1), scores, yerr=errs,
                marker='.', linestyle='-', lw=1, alpha=0.75,
                label=selector_label[idx])
plt.xlabel('Feature subset size')
plt.ylabel('Accurary')
plt.title('XGBoost')
plt.grid()
plt.legend(title='Selection algorithm', loc='lower right')
plt.show()


    

    




/home/jbourbeau/.local/lib/python2.7/site-packages/matplotlib/figure.py:1742: UserWarning: This figure includes Axes that are not compatible with tight_layout, so its results might be incorrect.
  warnings.warn("This figure includes Axes that are not "






    

In [15]:

    

k_features = 8
for idx, selector in enumerate(selector_list):
    indices_str = selector['feature_idx'][k_features-1]
    indices_str = indices_str.replace('(','')
    indices_str = indices_str.replace(')','')
    indices_str = indices_str.replace(',','')
    indices = [int(i) for i in indices_str.split()]
    print(selector_label[idx]+': (k = {})'.format(k_features))
    indices = [int(i) for i in indices_str.split()]
    print(np.sort(feature_labels[list(indices)]))


    

    




SFS: (k = 8)
['$\\cos(\\theta_{\\mathrm{Lap}})$' '$\\log_{10}$(NChannels (top 50\\%))'
 '$\\log_{10}$(NHits (top 50\\%))' '$\\log_{10}(S_{\\mathrm{500}})$'
 '$\\log_{10}(S_{\\mathrm{50}})$' '$r\\log_{10}(l)$'
 'Num HE stochastics (standard)' 'dE/dX (standard)']
SFFS: (k = 8)
['$\\cos(\\theta_{\\mathrm{Lap}})$' '$\\log_{10}$(NChannels (top 50\\%))'
 '$\\log_{10}$(NHits (top 50\\%))' '$\\log_{10}(S_{\\mathrm{500}})$'
 '$\\log_{10}(S_{\\mathrm{50}})$' '$r\\log_{10}(l)$'
 'Num HE stochastics (standard)' 'dE/dX (standard)']
SBS: (k = 8)
['$\\cos(\\theta_{\\mathrm{Lap}})$' '$\\log_{10}$(NChannels (top 50\\%))'
 '$\\log_{10}(E_{\\mathrm{Lap}}/\\mathrm{GeV})$'
 '$\\log_{10}(S_{\\mathrm{50}})$' '$r\\log_{10}(l)$' 'NChannels/NHits'
 'Num HE stochastics (standard)' 'dE/dX (standard)']
SFBS: (k = 8)
['$\\cos(\\theta_{\\mathrm{Lap}})$' '$\\log_{10}$(NChannels (top 50\\%))'
 '$\\log_{10}(E_{\\mathrm{Lap}}/\\mathrm{GeV})$'
 '$\\log_{10}(S_{\\mathrm{50}})$' '$r\\log_{10}(l)$' 'NChannels/NHits'
 'Num HE stochastics (standard)' 'dE/dX (standard)']

In [ ]:

    




    

In [ ]: