204-6-6-practice-non-linear-classifier-in-python

Link to the previous post : https://statinfer.com/204-6-5-the-non-linear-decision-boundary/

In this session we will practice non linear kernels of SVM in python.

Practice : Kernel – Non linear classifier

Dataset : Software users/sw_user_profile.csv
How many variables are there in software user profile data?
Plot the active users against and check weather the relation between age and “Active” status is linear or non-linear.
Build an SVM model(model-1), make sure that there is no kernel or the kernel is linear.
For model-1, create the confusion matrix and find out the accuracy.
Create a new variable. By using the polynomial kernel.
Build an SVM model(model-2), with the new data mapped on to higher dimensions. Keep the default kernel as linear.
For model-2, create the confusion matrix and find out the accuracy.
Plot the SVM with results.
With the original data re-cerate the model(model-3) and let python choose the default kernel function.
What is the accuracy of model-3?

In [19]:

#Dataset : Software users/sw_user_profile.csv  
sw_user_profile = pd.read_csv("datasets/Software users/sw_user_profile.csv")

In [20]:

#How many variables are there in software user profile data?
sw_user_profile.shape

Out[20]:

(490, 3)

In [21]:

#Plot the active users against and check weather the relation between age and "Active" status is linear or non-linear
plt.scatter(sw_user_profile.Age,sw_user_profile.Id,color='blue')

Out[21]:

<matplotlib.collections.PathCollection at 0xce7ac50>

In [22]:

#Build an SVM model(model-1), make sure that there is no kernel or the kernel is linear

#Model Building 
X= sw_user_profile[['Age']]
y= sw_user_profile[['Active']].values.ravel()
Linsvc = svm.SVC(kernel='linear', C=1).fit(X, y)

In [23]:

#Predicting values
predict3 = Linsvc.predict(X)

In [27]:

#For model-1, create the confusion matrix and find out the accuracy
#Confusion Matrix
from sklearn.metrics import confusion_matrix
conf_mat = confusion_matrix(sw_user_profile[['Active']],predict3)
conf_mat

Out[27]:

array([[317,   0],
       [173,   0]])

In [28]:

#Accuracy 
Accuracy3 = Linsvc.score(X, y)
Accuracy3

Out[28]:

0.64693877551020407

New variable derivation. Mapping to higher dimensions

In [29]:

#Standardizing the data to visualize the results clearly
sw_user_profile['age_nor']=(sw_user_profile.Age-numpy.mean(sw_user_profile.Age))/numpy.std(sw_user_profile.Age)

In [30]:

#Create a new variable. By using the polynomial kernel
#Creating the new variable
sw_user_profile['new']=(sw_user_profile.age_nor)*(sw_user_profile.age_nor)

In [31]:

#Build an SVM model(model-2), with the new data mapped on to higher dimensions. Keep the default kernel as linear

#Model Building with new variable
X= sw_user_profile[['Age']+['new']]
y= sw_user_profile[['Active']].values.ravel()
Linsvc = svm.SVC(kernel='linear', C=1).fit(X, y)
predict4 = Linsvc.predict(X)

In [32]:

#For model-2, create the confusion matrix and find out the accuracy
#Confusion Matrix
conf_mat = confusion_matrix(sw_user_profile[['Active']],predict4)
conf_mat

Out[32]:

array([[317,   0],
       [  0, 173]])

In [33]:

#Accuracy 
Accuracy4 = Linsvc.score(X, y)
Accuracy4

Out[33]:

1.0

In [34]:

#With the original data re-cerate the model(model-3) and let python choose the default kernel function. 
########Model Building with radial kernel function
X= sw_user_profile[['Age']]
y= sw_user_profile[['Active']].values.ravel()
Linsvc = svm.SVC(kernel='rbf', C=1).fit(X, y)
predict5 = Linsvc.predict(X)
conf_mat = confusion_matrix(sw_user_profile[['Active']],predict5)
conf_mat

Out[34]:

array([[317,   0],
       [  0, 173]])

In [35]:

#Accuracy model-3
Accuracy5 = Linsvc.score(X, y)
Accuracy5

Out[35]:

1.0

The next post is about soft margin classification noisy data.
Link to the next post : https://statinfer.com/204-6-7-soft-margin-classification-noisy-data/

13th March 2017