Linear Regression Model

Decision Tree Model

the accuracy, recall, precision comparsions among models

Other Metrics

We can see from the graph that at the recall value shown by the stright line, LogisticReg Model is higher value than NavieBayes model. So Logistic is the better model to predict the default as compare to NaiveBayes

From the above graph, we can see that we can adjust the Recall to be better that Percision by changing the classification threshold. ### Classifier with threshold of 0.2

Logistic Regression Model Instanization and Training

# Import the sklearn ML library - Logistic Regression
from sklearn.linear_model import LogisticRegression
# Instance of the estimator
logistic_regression = LogisticRegression(n_jobs=-1, random_state =15)
# train the estimator
logistic_regression.fit(X_train, y_train)

# Evaluate the model
y_pred_test = logistic_regression.predict(X_test)
# we conduct our evaluation using these 3 _score
eval_metrix.loc['accuracy', 'LogisticReg Model']= accuracy_score(y_pred=y_pred_test, y_true=y_test)
eval_metrix.loc['precision', 'LogisticReg Model']= precision_score(y_pred=y_pred_test, y_true=y_test)
eval_metrix.loc['recall', 'LogisticReg Model']= recall_score(y_pred=y_pred_test, y_true=y_test)    

# Confusion matrix for the logistic regression
CM = confusion_matrix(y_pred=y_pred_test, y_true=y_test)
CMatrix(CM)