The file "loanacceptance.csv" contains various attributes of 500 customers based on which loan has either been granted or denied. You have to create a system that automatically decides whether to grant a loan or not to grant a loan.

Importing libraries

In [15]:
from sklearn import svm
from sklearn import preprocessing
from sklearn import metrics
import numpy as np
import csv

Reading data file as list object

In [16]:
def readFileThroughCSV(filename):
    csvfile = open(filename)

    # creating a csv reader object
    readerobject = csv.reader(csvfile, delimiter=',')
    lst = list(readerobject)
    csvfile.close()

    # removing first row from list
    lst = lst[1:]
    arr = np.array(lst)
    data = arr.astype(float)

    # extract last column which is classification label
    c = data[:,-1]
    # extract remaining data
    d = data[:,1:-1]
    
    return(c,d)
In [31]:
import pandas as pd
def readFileThroughPandas(filename):
    
    # Reads the entire data file
    data = pd.read_csv(filename)
    
    # Following command will read the columns 1 to 7 leaving out the 0th column
    # data = pd.read_csv(filename, usecols = np.arange(1,8))
    
    c = data["Loan Granted"]
    d = data[["Marital Status","Kids","Annual Household Salary","Loan Amount","Car owner", "Education Level"]]
    
    c = c.to_numpy()
    d = d.to_numpy()
   
    return(c,d)
In [33]:
(c,d) = readFileThroughPandas("loanacceptance.csv")
# shape of the variables
print(c.shape)
print(d.shape)

print(type(c))
print(type(d))
(500,)
(500, 6)
<class 'numpy.ndarray'>
<class 'numpy.ndarray'>

Fitting an SVM model

In [27]:
# Create an SVM classification object

# clf = svm.SVC(kernel='linear',C=1)
# Note that rbf is the default kernel in svm.SVC
# C is a regularization parameter to avoid overfitting

clf = svm.SVC(kernel='linear',C=1)

# Let us do a 0-1 scaling of our dataset because the attributes are of significantly different orders of magnitude
d = preprocessing.scale(d)

# Fitting SVM only on first 400 data points
clf.fit(X=d[0:400,:],y=c[0:400])

# In case of more than 2 classes, note that multiclass is done based on one-vs-one in svm.SVC
Out[27]:
SVC(C=1, kernel='linear')

Printing performance metrics

In [28]:
# returns accuracy
print("Training accuracy",clf.score(X=d[:400,:],y=c[:400]))
print("Testing accuracy",clf.score(X=d[400:,:],y=c[400:]))
Training accuracy 0.935
Testing accuracy 0.88

Always look at the confusion matrix

In [29]:
indices = range(400,500)
c_predicted = clf.predict(d[indices,:])
m = metrics.confusion_matrix(c[indices],c_predicted)
print(m)
[[24  6]
 [ 6 64]]