notebook.community

Edit and run 





In [61]:



    


import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sb
from operator import itemgetter
%matplotlib inline



    











In [3]:



    


results = {'eval': {'auc': [0.499306, 0.494721, 0.488865, 0.478342, 0.5674]},
 'train': {'auc': [0.557143, 0.632473, 0.715275, 0.74478, 0.780549]}}

labels = ['max_depth','eta','num_rounds','eval_reslts']
data = [[3,0.1,100,results],[3,0.1,100,results],[3,0.1,100,results]]
df = pd.DataFrame(data=data,columns=labels)



    











In [4]:



    


df



    


















    
Out[4]:










  
    

      
      max_depth
      eta
      num_rounds
      eval_reslts
    

  
  
    

      0
      3
      0.1
      100
      {u'train': {u'auc': [0.557143, 0.632473, 0.715...
    

    

      1
      3
      0.1
      100
      {u'train': {u'auc': [0.557143, 0.632473, 0.715...
    

    

      2
      3
      0.1
      100
      {u'train': {u'auc': [0.557143, 0.632473, 0.715...
    

  




















In [2]:



    


def myfunc(x):
    
    val = np.array(x['eval']['auc'])
    maxInd = np.argmax(val)
    maxVal = val[maxInd]
    
    return (maxInd,maxVal)



    











In [21]:



    


test = df['eval_reslts'].loc[0]
print test['eval'].values



    


















    






<built-in method values of dict object at 0x7fc0bbb6d168>

















In [22]:



    


test = df['eval_reslts'].loc[0]
myfunc(test)



    


















    
Out[22]:








(4, 0.56740000000000002)

















In [24]:



    


df['eval_auc'] = df['eval_reslts'].map(lambda x: myfunc(x))



    











In [25]:



    


df



    


















    
Out[25]:










  
    

      
      max_depth
      eta
      num_rounds
      eval_reslts
      eval_auc
    

  
  
    

      0
      3
      0.1
      100
      {u'train': {u'auc': [0.557143, 0.632473, 0.715...
      (4, 0.5674)
    

    

      1
      3
      0.1
      100
      {u'train': {u'auc': [0.557143, 0.632473, 0.715...
      (4, 0.5674)
    

    

      2
      3
      0.1
      100
      {u'train': {u'auc': [0.557143, 0.632473, 0.715...
      (4, 0.5674)
    

  




















In [2]:



    


def build_roc(df):
    
    df['TPR'] = df['recall']
    df['FPR'] = df['FP']/(df['FP'] + df['TN'])
    
#     plt.plot([0,1],[0,1],'k',linewidth=0.5)
    plt.figure()
    plt.plot(df.FPR.values,df.TPR.values,'r*',markersize=7)
    plt.xlabel('FPR')
    plt.xlim([0,1])
    plt.ylabel('TPR')
    plt.ylim([0,1])
    titlestr = "AUC: {}  k = {}".format(np.trapz(df.TPR.values[::-1],x=df.FPR.values[::-1]),int(df.k.unique()))
    plt.title(titlestr)



    











In [ ]:



    


for k in df.k.unique():
    build_roc(df[df['k']==k])



    











In [18]:



    


path = '../../data/gridsearch_xgb.csv'



    











In [19]:



    


df = pd.read_csv(path)



    











In [20]:



    


df.head()



    


















    
Out[20]:










  
    

      
      Unnamed: 0
      num_rounds
      max_depth
      eta
      eval_results
    

  
  
    

      0
      0
      100
      3
      0.03
      {'train': {'auc': ['0.552844', '0.552849', '0....
    

    

      1
      1
      300
      3
      0.03
      {'train': {'auc': ['0.552844', '0.552849', '0....
    

    

      2
      2
      600
      3
      0.03
      {'train': {'auc': ['0.552844', '0.552849', '0....
    

    

      3
      3
      900
      3
      0.03
      {'train': {'auc': ['0.552844', '0.552849', '0....
    

    

      4
      4
      100
      4
      0.03
      {'train': {'auc': ['0.564531', '0.564540', '0....
    

  




















In [23]:



    


df['eval_results'].ix[0]



    


















    
Out[23]:








"{'train': {'auc': ['0.552844', '0.552849', '0.552853', '0.552854', '0.564506', '0.564508', '0.564585', '0.570304', '0.570369', '0.570379', '0.574745', '0.574746', '0.574748', '0.574748', '0.574755', '0.574814', '0.577942', '0.577889', '0.577944', '0.591977', '0.591976', '0.591979', '0.595014', '0.605616', '0.609023', '0.609023', '0.609026', '0.609065', '0.615975', '0.616123', '0.616305', '0.616303', '0.616452', '0.616523', '0.616298', '0.616303', '0.648427', '0.648950', '0.648568', '0.652931', '0.652968', '0.652983', '0.653288', '0.653320', '0.657873', '0.657791', '0.657966', '0.657924', '0.658021', '0.658081', '0.658085', '0.658091', '0.663862', '0.663871', '0.663874', '0.663901', '0.665859', '0.668477', '0.668617', '0.668639', '0.668658', '0.668621', '0.668623', '0.670125', '0.670147', '0.670208', '0.669043', '0.671257', '0.671260', '0.677831', '0.680031', '0.680049', '0.680268', '0.680302', '0.680307', '0.680937', '0.680934', '0.682168', '0.682185', '0.682234', '0.682249', '0.683827', '0.685565', '0.686696', '0.686442', '0.687807', '0.687826', '0.687395', '0.686994', '0.687051', '0.687506', '0.690304', '0.690345', '0.690778', '0.690929', '0.691567', '0.691241', '0.696854', '0.696816', '0.697191']}, 'eval': {'auc': ['0.58177', '0.58177', '0.58177', '0.58177', '0.57943', '0.57943', '0.57943', '0.57936', '0.57936', '0.57936', '0.57965', '0.57965', '0.57965', '0.57965', '0.57965', '0.57966', '0.57966', '0.57965', '0.57966', '0.67961', '0.67961', '0.67961', '0.67961', '0.68028', '0.68028', '0.68027', '0.68027', '0.68028', '0.68180', '0.68180', '0.68180', '0.68180', '0.68220', '0.68220', '0.68180', '0.68180', '0.70203', '0.70203', '0.70203', '0.70193', '0.70204', '0.70204', '0.70204', '0.70204', '0.70215', '0.70214', '0.70215', '0.70203', '0.70193', '0.70204', '0.70204', '0.70204', '0.69895', '0.69895', '0.69894', '0.69894', '0.69894', '0.69910', '0.69943', '0.69944', '0.69944', '0.69944', '0.69944', '0.69927', '0.69928', '0.69929', '0.69931', '0.69916', '0.69916', '0.69586', '0.69586', '0.69588', '0.69584', '0.69584', '0.69584', '0.69512', '0.69513', '0.69506', '0.69509', '0.69509', '0.69509', '0.69585', '0.69584', '0.69583', '0.69561', '0.69560', '0.69560', '0.69548', '0.69547', '0.69547', '0.69557', '0.70430', '0.70409', '0.70395', '0.70395', '0.70396', '0.70396', '0.70401', '0.70427', '0.70424']}}"

















In [66]:



    


def findtrain(x):
    badset = {"{","}","'",",","t","r","a","i","n","u","c",":","[","]"}
    outstr = ""
    results = x.split("eval")
    
    for char in results[0]:
        if char not in badset:
            outstr = outstr + char
            
    numbers = outstr.split(" ")
    aucVals = np.array([float(number) for number in numbers[2:-1]])
    maxInd = np.argmax(aucVals)
    maxVal = aucVals[maxInd]
    
    return (maxInd,maxVal)
  
def findbest(x):
    numvals = len(x)
    bestrow = None
    bestVal = None
    
    for row in xrange(numvals):
        if x[row][1] > bestVal:
            bestrow = row
            bestVal = x[row][1]
            
    return bestrow, x[bestrow]


def findeval(x):
    badset = {"{","}","'",",","t","r","a","i","n","u","c",":","[","]"}
    outstr = ""
    results = x.split("eval")
    
    for char in results[1]:
        if char not in badset:
            outstr = outstr + char
            
    numbers = outstr.split(" ")
    aucVals = np.array([float(number) for number in numbers[2:]])
    maxInd = np.argmax(aucVals)
    maxVal = aucVals[maxInd]
    
    return (maxInd,maxVal)



    











In [57]:



    


df['eval_auc'] = df['eval_results'].map(lambda x: findeval(x))



    











In [58]:



    


df['train_auc'] = df['eval_results'].map(lambda x: findtrain(x))



    











In [60]:



    


df



    


















    
Out[60]:










  
    

      
      Unnamed: 0
      num_rounds
      max_depth
      eta
      eval_results
      eval_auc
      train_auc
    

  
  
    

      0
      0
      100
      3
      0.03
      {'train': {'auc': ['0.552844', '0.552849', '0....
      (91, 0.7043)
      (99, 0.697191)
    

    

      1
      1
      300
      3
      0.03
      {'train': {'auc': ['0.552844', '0.552849', '0....
      (284, 0.74588)
      (299, 0.743419)
    

    

      2
      2
      600
      3
      0.03
      {'train': {'auc': ['0.552844', '0.552849', '0....
      (596, 0.80778)
      (599, 0.767899)
    

    

      3
      3
      900
      3
      0.03
      {'train': {'auc': ['0.552844', '0.552849', '0....
      (661, 0.80968)
      (897, 0.782626)
    

    

      4
      4
      100
      4
      0.03
      {'train': {'auc': ['0.564531', '0.564540', '0....
      (97, 0.70845)
      (98, 0.708373)
    

    

      5
      5
      300
      4
      0.03
      {'train': {'auc': ['0.564531', '0.564540', '0....
      (292, 0.77796)
      (299, 0.753988)
    

    

      6
      6
      600
      4
      0.03
      {'train': {'auc': ['0.564531', '0.564540', '0....
      (522, 0.81082)
      (599, 0.77964)
    

    

      7
      7
      900
      4
      0.03
      {'train': {'auc': ['0.564531', '0.564540', '0....
      (899, 0.81823)
      (899, 0.792982)
    

    

      8
      8
      100
      5
      0.03
      {'train': {'auc': ['0.570415', '0.570409', '0....
      (91, 0.71407)
      (99, 0.718837)
    

    

      9
      9
      300
      5
      0.03
      {'train': {'auc': ['0.570415', '0.570409', '0....
      (280, 0.79814)
      (299, 0.76341)
    

    

      10
      10
      600
      5
      0.03
      {'train': {'auc': ['0.570415', '0.570409', '0....
      (388, 0.81023)
      (599, 0.788445)
    

    

      11
      11
      900
      5
      0.03
      {'train': {'auc': ['0.570415', '0.570409', '0....
      (881, 0.82387)
      (899, 0.802094)
    

    

      12
      12
      100
      6
      0.03
      {'train': {'auc': ['0.574951', '0.574936', '0....
      (99, 0.72097)
      (97, 0.727374)
    

    

      13
      13
      300
      6
      0.03
      {'train': {'auc': ['0.574951', '0.574936', '0....
      (290, 0.80857)
      (299, 0.771393)
    

    

      14
      14
      600
      6
      0.03
      {'train': {'auc': ['0.574951', '0.574936', '0....
      (596, 0.81735)
      (599, 0.795562)
    

    

      15
      15
      900
      6
      0.03
      {'train': {'auc': ['0.574951', '0.574936', '0....
      (837, 0.82755)
      (899, 0.809408)
    

    

      16
      16
      100
      3
      0.06
      {'train': {'auc': ['0.552844', '0.552849', '0....
      (98, 0.73037)
      (99, 0.725216)
    

    

      17
      17
      300
      3
      0.06
      {'train': {'auc': ['0.552844', '0.552849', '0....
      (286, 0.80978)
      (299, 0.76855)
    

    

      18
      18
      600
      3
      0.06
      {'train': {'auc': ['0.552844', '0.552849', '0....
      (583, 0.81891)
      (599, 0.792096)
    

    

      19
      19
      900
      3
      0.06
      {'train': {'auc': ['0.552844', '0.552849', '0....
      (886, 0.82937)
      (899, 0.805351)
    

    

      20
      20
      100
      4
      0.06
      {'train': {'auc': ['0.564531', '0.570346', '0....
      (95, 0.74017)
      (99, 0.739443)
    

    

      21
      21
      300
      4
      0.06
      {'train': {'auc': ['0.564531', '0.570346', '0....
      (257, 0.81035)
      (299, 0.780313)
    

    

      22
      22
      600
      4
      0.06
      {'train': {'auc': ['0.564531', '0.570346', '0....
      (534, 0.8261)
      (599, 0.803442)
    

    

      23
      23
      900
      4
      0.06
      {'train': {'auc': ['0.564531', '0.570346', '0....
      (839, 0.84644)
      (899, 0.815612)
    

    

      24
      24
      100
      5
      0.06
      {'train': {'auc': ['0.570415', '0.570412', '0....
      (92, 0.74551)
      (99, 0.748918)
    

    

      25
      25
      300
      5
      0.06
      {'train': {'auc': ['0.570415', '0.570412', '0....
      (205, 0.81014)
      (299, 0.789367)
    

    

      26
      26
      600
      5
      0.06
      {'train': {'auc': ['0.570415', '0.570412', '0....
      (599, 0.8387)
      (599, 0.812053)
    

    

      27
      27
      900
      5
      0.06
      {'train': {'auc': ['0.570415', '0.570412', '0....
      (892, 0.84947)
      (899, 0.823539)
    

    

      28
      28
      100
      6
      0.06
      {'train': {'auc': ['0.574951', '0.613902', '0....
      (78, 0.74694)
      (99, 0.755981)
    

    

      29
      29
      300
      6
      0.06
      {'train': {'auc': ['0.574951', '0.613902', '0....
      (299, 0.82282)
      (299, 0.795567)
    

    

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

    

      66
      66
      600
      3
      0.60
      {'train': {'auc': ['0.552844', '0.570280', '0....
      (152, 0.85972)
      (599, 0.851604)
    

    

      67
      67
      900
      3
      0.60
      {'train': {'auc': ['0.552844', '0.570280', '0....
      (152, 0.85972)
      (899, 0.86043)
    

    

      68
      68
      100
      4
      0.60
      {'train': {'auc': ['0.564531', '0.613752', '0....
      (71, 0.83969)
      (99, 0.81354)
    

    

      69
      69
      300
      4
      0.60
      {'train': {'auc': ['0.564531', '0.613752', '0....
      (235, 0.85638)
      (299, 0.844328)
    

    

      70
      70
      600
      4
      0.60
      {'train': {'auc': ['0.564531', '0.613752', '0....
      (235, 0.85638)
      (599, 0.86115)
    

    

      71
      71
      900
      4
      0.60
      {'train': {'auc': ['0.564531', '0.613752', '0....
      (235, 0.85638)
      (899, 0.870383)
    

    

      72
      72
      100
      5
      0.60
      {'train': {'auc': ['0.570415', '0.603725', '0....
      (93, 0.85893)
      (99, 0.820764)
    

    

      73
      73
      300
      5
      0.60
      {'train': {'auc': ['0.570415', '0.603725', '0....
      (93, 0.85893)
      (299, 0.852038)
    

    

      74
      74
      600
      5
      0.60
      {'train': {'auc': ['0.570415', '0.603725', '0....
      (93, 0.85893)
      (599, 0.869036)
    

    

      75
      75
      900
      5
      0.60
      {'train': {'auc': ['0.570415', '0.603725', '0....
      (93, 0.85893)
      (899, 0.879071)
    

    

      76
      76
      100
      6
      0.60
      {'train': {'auc': ['0.574951', '0.644546', '0....
      (74, 0.84766)
      (99, 0.828738)
    

    

      77
      77
      300
      6
      0.60
      {'train': {'auc': ['0.574951', '0.644546', '0....
      (131, 0.85167)
      (299, 0.859913)
    

    

      78
      78
      600
      6
      0.60
      {'train': {'auc': ['0.574951', '0.644546', '0....
      (131, 0.85167)
      (599, 0.877316)
    

    

      79
      79
      900
      6
      0.60
      {'train': {'auc': ['0.574951', '0.644546', '0....
      (131, 0.85167)
      (899, 0.88754)
    

    

      80
      80
      100
      3
      0.90
      {'train': {'auc': ['0.552844', '0.574873', '0....
      (94, 0.8327)
      (99, 0.805588)
    

    

      81
      81
      300
      3
      0.90
      {'train': {'auc': ['0.552844', '0.574873', '0....
      (138, 0.85149)
      (299, 0.839067)
    

    

      82
      82
      600
      3
      0.90
      {'train': {'auc': ['0.552844', '0.574873', '0....
      (138, 0.85149)
      (599, 0.856192)
    

    

      83
      83
      900
      3
      0.90
      {'train': {'auc': ['0.552844', '0.574873', '0....
      (138, 0.85149)
      (899, 0.865057)
    

    

      84
      84
      100
      4
      0.90
      {'train': {'auc': ['0.564531', '0.626056', '0....
      (95, 0.83787)
      (99, 0.816703)
    

    

      85
      85
      300
      4
      0.90
      {'train': {'auc': ['0.564531', '0.626056', '0....
      (134, 0.83875)
      (299, 0.849048)
    

    

      86
      86
      600
      4
      0.90
      {'train': {'auc': ['0.564531', '0.626056', '0....
      (134, 0.83875)
      (599, 0.866013)
    

    

      87
      87
      900
      4
      0.90
      {'train': {'auc': ['0.564531', '0.626056', '0....
      (134, 0.83875)
      (899, 0.875472)
    

    

      88
      88
      100
      5
      0.90
      {'train': {'auc': ['0.570415', '0.637539', '0....
      (74, 0.84632)
      (99, 0.825646)
    

    

      89
      89
      300
      5
      0.90
      {'train': {'auc': ['0.570415', '0.637539', '0....
      (113, 0.84751)
      (299, 0.857447)
    

    

      90
      90
      600
      5
      0.90
      {'train': {'auc': ['0.570415', '0.637539', '0....
      (113, 0.84751)
      (599, 0.874347)
    

    

      91
      91
      900
      5
      0.90
      {'train': {'auc': ['0.570415', '0.637539', '0....
      (113, 0.84751)
      (899, 0.884148)
    

    

      92
      92
      100
      6
      0.90
      {'train': {'auc': ['0.574951', '0.645773', '0....
      (85, 0.84649)
      (99, 0.833498)
    

    

      93
      93
      300
      6
      0.90
      {'train': {'auc': ['0.574951', '0.645773', '0....
      (85, 0.84649)
      (299, 0.864634)
    

    

      94
      94
      600
      6
      0.90
      {'train': {'auc': ['0.574951', '0.645773', '0....
      (85, 0.84649)
      (599, 0.882257)
    

    

      95
      95
      900
      6
      0.90
      {'train': {'auc': ['0.574951', '0.645773', '0....
      (85, 0.84649)
      (899, 0.892832)
    

  



96 rows × 7 columns



















In [63]:



    


max(df['eval_auc'].values,key=itemgetter(1))



    


















    
Out[63]:








(163, 0.86024999999999996)

















In [67]:



    


findbest(df['eval_auc'].values)



    


















    
Out[67]:








(53, (163, 0.86024999999999996))

















In [68]:



    


df.ix[53]



    


















    
Out[68]:








Unnamed: 0                                                     53
num_rounds                                                    300
max_depth                                                       4
eta                                                           0.3
eval_results    {'train': {'auc': ['0.564531', '0.574791', '0....
eval_auc                                           (163, 0.86025)
train_auc                                         (299, 0.829813)
Name: 53, dtype: object

















In [ ]:



    






    











In [34]:



    


results = x.split("eval")



    











In [51]:



    


results[0]



    


















    
Out[51]:








"{'train': {'auc': ['0.552844', '0.552849', '0.552853', '0.552854', '0.564506', '0.564508', '0.564585', '0.570304', '0.570369', '0.570379', '0.574745', '0.574746', '0.574748', '0.574748', '0.574755', '0.574814', '0.577942', '0.577889', '0.577944', '0.591977', '0.591976', '0.591979', '0.595014', '0.605616', '0.609023', '0.609023', '0.609026', '0.609065', '0.615975', '0.616123', '0.616305', '0.616303', '0.616452', '0.616523', '0.616298', '0.616303', '0.648427', '0.648950', '0.648568', '0.652931', '0.652968', '0.652983', '0.653288', '0.653320', '0.657873', '0.657791', '0.657966', '0.657924', '0.658021', '0.658081', '0.658085', '0.658091', '0.663862', '0.663871', '0.663874', '0.663901', '0.665859', '0.668477', '0.668617', '0.668639', '0.668658', '0.668621', '0.668623', '0.670125', '0.670147', '0.670208', '0.669043', '0.671257', '0.671260', '0.677831', '0.680031', '0.680049', '0.680268', '0.680302', '0.680307', '0.680937', '0.680934', '0.682168', '0.682185', '0.682234', '0.682249', '0.683827', '0.685565', '0.686696', '0.686442', '0.687807', '0.687826', '0.687395', '0.686994', '0.687051', '0.687506', '0.690304', '0.690345', '0.690778', '0.690929', '0.691567', '0.691241', '0.696854', '0.696816', '0.697191']}, '"

















In [54]:



    


outstr = ""
badset = {"{","}","'",",","t","r","a","i","n","u","c",":","[","]"}
for char in results[0]:
        if char not in badset:
            outstr = outstr + char



    











In [55]:



    


outstr.split(" ")



    


















    
Out[55]:








['',
 '',
 '0.552844',
 '0.552849',
 '0.552853',
 '0.552854',
 '0.564506',
 '0.564508',
 '0.564585',
 '0.570304',
 '0.570369',
 '0.570379',
 '0.574745',
 '0.574746',
 '0.574748',
 '0.574748',
 '0.574755',
 '0.574814',
 '0.577942',
 '0.577889',
 '0.577944',
 '0.591977',
 '0.591976',
 '0.591979',
 '0.595014',
 '0.605616',
 '0.609023',
 '0.609023',
 '0.609026',
 '0.609065',
 '0.615975',
 '0.616123',
 '0.616305',
 '0.616303',
 '0.616452',
 '0.616523',
 '0.616298',
 '0.616303',
 '0.648427',
 '0.648950',
 '0.648568',
 '0.652931',
 '0.652968',
 '0.652983',
 '0.653288',
 '0.653320',
 '0.657873',
 '0.657791',
 '0.657966',
 '0.657924',
 '0.658021',
 '0.658081',
 '0.658085',
 '0.658091',
 '0.663862',
 '0.663871',
 '0.663874',
 '0.663901',
 '0.665859',
 '0.668477',
 '0.668617',
 '0.668639',
 '0.668658',
 '0.668621',
 '0.668623',
 '0.670125',
 '0.670147',
 '0.670208',
 '0.669043',
 '0.671257',
 '0.671260',
 '0.677831',
 '0.680031',
 '0.680049',
 '0.680268',
 '0.680302',
 '0.680307',
 '0.680937',
 '0.680934',
 '0.682168',
 '0.682185',
 '0.682234',
 '0.682249',
 '0.683827',
 '0.685565',
 '0.686696',
 '0.686442',
 '0.687807',
 '0.687826',
 '0.687395',
 '0.686994',
 '0.687051',
 '0.687506',
 '0.690304',
 '0.690345',
 '0.690778',
 '0.690929',
 '0.691567',
 '0.691241',
 '0.696854',
 '0.696816',
 '0.697191',
 '']

















In [ ]:

Content source: mgupta011235/TweetSafe

Similar notebooks:

notebook.community | gallery | about