Logistic regression with the ability of implementing penalty terms. The class object is dedicated to binary classification and is compatible with scikit-learn's GridSearchCV method.
An example of its usage was performed on the WBCD dataset and returned the following estimates:
- Accuracy: 0.9883
- Recall: 0.9841
- Precision: 0.9841
- Error Rate: 0.0117
- F1 Score: 0.9841
- ROC: 0.9874
- Specificity: 0.9907
- Misclassified Samples: 2 (out of 171)
Code example:
from logistic_regression import LogisticRegression
log_model = LogisticRegression(learning_rate=0.01, C=0.1, num_iter=20, penalty='elasticnet', l1_ratio=0.7)
log_model.fit(X_train, y_train)
predictions = log_model.predict(X_test, threshold=0.6)