scoring 0.0.9

Build scorecard for credit risk analysis

import pandas as pd
import numpy as np
import scoring as sc

from sklearn.model_selection import train_test_split as tts
from sklearn.linear_model import LogisticRegression as lr
import sklearn.metrics as metrics

df=pd.read_csv('gc.csv')
vardict=pd.read_csv('dict.csv')
df['Risk']=df['Risk'].apply(lambda x: 1 if x=='bad' else 0)
df=sc.renameCols(df,vardict,False)
label,disc,cont=sc.getVarTypes(vardict)
# sc.discSummary(df)

# ### No row needs to be removed from this example in this stage ###
# vardict.loc[vardict['new'].isin(['Age','Sex']),'isDel']=1
# df,vardict=cl.delFromVardict(df,vardict)

df1=sc.binData(df,vardict)

#########################################
####It's using Chi-Merge algorithm...####
#########################################

Doing continous feature: Age

Doing continous feature: Credit amount
Equal Depth Binning is required, number of bins is: 100

Doing continous feature: Duration

Doing discrete feature: Sex

Doing discrete feature: Job

Doing discrete feature: Housing

Doing discrete feature: Saving accounts

Doing discrete feature: Checking account

Doing discrete feature: Purpose

Finished

bidict=sc.getBiDict(df1,label)

bidict['Credit amount']

<style scoped> .dataframe tbody tr th:only-of-type { vertical-align: middle; }

.dataframe tbody tr th {
    vertical-align: top;
}

.dataframe thead th {
    text-align: right;
}

</style>

	Credit amount	total	good	bad	totalDist	goodDist	badDist	goodRate	badRate	woe	iv
0	(-inf, 1282.0]	211	144	67	0.211	0.223	0.206	0.682	0.318	-0.082	0.001
1	(1282.0, 3446.32]	469	352	117	0.469	0.390	0.503	0.751	0.249	0.254	0.029
2	(3446.32, 3913.26]	60	55	5	0.060	0.017	0.079	0.917	0.083	1.551	0.096
3	(3913.26, inf]	260	149	111	0.260	0.370	0.213	0.573	0.427	-0.553	0.087

# modified credit amount
sc.bivariate(pd.DataFrame({'y':df['y'],
                           'Credit amount':sc.manuallyBin(df,
                                                          'Credit amount',
                                                          'cont',
                                                          [-np.inf,1300,3500,4000,np.inf])}
                         ),'Credit amount','y')[0]
df1['Credit amount']=sc.manuallyBin(df,'Credit amount','cont',[-np.inf,1300,3500,4000,np.inf])

bidict=sc.getBiDict(df1,label)
ivtable=sc.ivTable(bidict)

df1,vardict,bidict=sc.featureFilter(df1,vardict,bidict,ivtable)

df=sc.mapWOE(df1,bidict,label)

### Modelling ###
#################
trainx,testx,trainy,testy=tts(df.iloc[:,1:],df[label],test_size=0.3)
m=lr(penalty='l1', C=0.9, solver='saga', n_jobs=-1)
m.fit(trainx,trainy)
pred=m.predict(testx)
pred_prob=m.predict_proba(testx)[:,1]

# 鏌ョ湅娴嬭瘯缁撴灉
cm=metrics.confusion_matrix(testy, pred)
print('**Precision is:',(cm[0][0]+cm[1][1])/(sum(cm[0])+sum(cm[1])))
print('\n**Confusion matrix is:\n',cm)
print('\n**Classification report is:\n',metrics.classification_report(testy, pred))

**Precision is: 0.7233333333333334

**Confusion matrix is:
 [[179  18]
 [ 65  38]]

**Classification report is:
               precision    recall  f1-score   support

           0       0.73      0.91      0.81       197
           1       0.68      0.37      0.48       103

   micro avg       0.72      0.72      0.72       300
   macro avg       0.71      0.64      0.64       300
weighted avg       0.71      0.72      0.70       300

### Evaluation ###
##################
sc.plotROC(testy,pred_prob)
sc.plotKS(testy,pred_prob)
sc.plotCM(metrics.confusion_matrix(testy,pred), classes=df[label].unique(),
          title='Confusion matrix, without normalization')

Confusion matrix, without normalization
[[179  18]
 [ 65  38]]

### Scoring ###
###############
scored,basescore=sc.scoring(trainx.reset_index(drop=True),
                            trainy.reset_index(drop=True),
                            'y',
                            m,
                            bidict)