In [2]:

    

# 1. Import sklearn, import pandas as pd, and pd.read_csv the CFPB CSV file into dataframe 'df'.

import sklearn
import pandas as pd

df = pd.read_csv('data/cfpb_complaints_with_fictitious_data.csv')


    

In [3]:

    

# 2. Filter your df down to 'Product', 'Consumer Claim', 'Amount Received' using [[]] notation. Which is our target?

df = df[['Product', 'Consumer Claim', 'Amount Received']]
df.head(5) # Our target is "Product"


    

    
Out[3]:






  

    

      

      
Product
      
Consumer Claim
      
Amount Received
    
  
  

    

      
0
      
Credit card
      
332.63
      
130.22
    
    

      
1
      
Debt collection
      
54.79
      
49.14
    
    

      
2
      
Credit card
      
215.04
      
155.28
    
    

      
3
      
Credit reporting
      
3.31
      
4.59
    
    

      
4
      
Debt collection
      
73.99
      
46.02
    
  

In [6]:

    

# 3. From sklearn.cross_validation import train_test_split. Make a train/test split 80/20 (we won't use it though).

from sklearn.cross_validation import train_test_split

train, test = train_test_split(df, test_size=.2)
print(len(train))
print(len(test))


    

    




100544
25136

In [7]:

    

# 4. Assign df[['Consumer Claim', 'Amount Received']] to 'X'

X = df[['Consumer Claim', 'Amount Received']]


    

In [8]:

    

# 5. Convert to raw values df['Product'].values and assign to 'y'

y = df['Product']


    

In [9]:

    

# 6. From sklearn.preprocessing import StandardScaler. From sklearn.pipeline import Pipeline.

from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline


    

In [11]:

    

# 7. From sklearn.neighbors import KNeighborsClassifier. Make a scalar/knn pipeline.

from sklearn.neighbors import KNeighborsClassifier

pipe = Pipeline([('scaler', StandardScaler()),
                 ('classifier', KNeighborsClassifier())])


    

In [12]:

    

# 8. Fit your pipeline with your X and y.

pipe.fit(X, y)


    

    
Out[12]:





Pipeline(steps=[('scaler', StandardScaler(copy=True, with_mean=True, with_std=True)), ('classifier', KNeighborsClassifier(algorithm='auto', leaf_size=30, metric='minkowski',
           metric_params=None, n_jobs=1, n_neighbors=5, p=2,
           weights='uniform'))])

In [13]:

    

# 9. Use your newly fitted pipeline to predict classifications for [[100, 80], [5000, 4000], [350, 900]] .

pipe.predict([[100, 80], [5000, 4000], [350, 900]])


    

    
Out[13]:





array(['Money transfers', 'Consumer Loan', 'Other financial service'], dtype=object)

In [16]:

    

# 10. From sklearn.cross_validation import cross_val_score. Run cross val score on your pipeline.

from sklearn.cross_validation import cross_val_score

scores = cross_val_score(pipe, X, y)
scores


    

    
Out[16]:





array([ 0.89779942,  0.89747691,  0.89510373])

In [15]:

    

# 11. Get the mean of cross validation scores from your pipeline,

score_mean = scores.mean()
score_mean


    

    
Out[15]:





0.89679334986607062

In [ ]:

    

# 12. Now repeat with Support Vector Machine Classifier (sklearn.svm.SVC) pipeline. Which yields better results?

from sklearn.svm import SVC

pipe = Pipeline([('scaler', StandardScaler()),
                 ('classifier', SVC())])

pipe.fit(X, y)

print(pipe.predict([[100, 80], [5000, 4000], [350, 900]]))

scores = cross_val_score(pipe, X, y)
print(scores)

print(scores.mean())


    

    




['Debt collection' 'Consumer Loan' 'Debt collection']