hyper_parameter_tuning.py

### Hyper parameter tuning Logistic Regression model

from sklearn.model_selection import GridSearchCV
from sklearn import linear_model
from sklearn.pipeline import make_pipeline

lr = linear_model.LogisticRegression(solver ='lbfgs', penalty = 'l2')

# Use gridsearch CV to search for the bes parameter

grid = GridSearchCV(lr, {'C':[0.0001,0.001,0.01,0.1,1,10]})
grid.fit(X_train, Y_train)

# Print out the best parameter
print("Optimal Regularization strength is :", grid.best_params_)

#Initializing logistic regression object
lr2 = linear_model.LogisticRegression( C = 1, penalty = 'l2')
lr2.fit(X_train, Y_train)
print("Accuracy score of the logistic regression model ",lr2.score(X_test, Y_test)*100,"%")

### Plot the inverse regularization strength
train_errors , test_errors = [], []
C_list = [0.0001,0.001,0.01,0.1,1,10]
for i in C_list:
  lr3 = linear_model.LogisticRegression( C = i, penalty = 'l2')
  lr3.fit(X_train, Y_train)

  ## Evaluate train and test error rates
  train_errors.append(lr3.score(X_train, Y_train))
  test_errors.append(lr3.score(X_test, Y_test))

plt.semilogx(C_list, train_errors,C_list, test_errors )
plt.legend(("train","test"))
plt.xlabel("Inverse regularization strength")
plt.ylabel('Accuracy score')
plt.show()