This is the Logistic Regression demo implement from Scratcth 😄 .

To install packages, install from requirements.txt file via this command:

pip install -r requirements.txt

You can get Logistic Regression class from the LogisticRegresion.py file, for example:

from LogisticRegression import LogisticRegression

X_train = ...
y_train = ...

X_val = ...
y_val = ...

num_iters = 1000

model = LogisticRegression(lr = 1e-2, lambda_val = 1e-3)

model.train(X_train, y_train, X_test, y_test, num_iters)

acc, precision, recall, f1_score = model.evaluate(X_test, y_test)

print(acc, precision, recall, f1_score)

All of the examples you can find in the train.py file.

You can get the sigmoid(z) from the file above.

Data sample can be found at the training_data.txt file

Model can be saved at the model.json file, which includes the weight of the model, you can use load_weight attribute from the model to load the model json file.

from LogisticRegression import LogisticRegression
model = LogisticRegression(lr = 1e-2, lambda_val = 1e-3)
model.load_weights("model.json")
model.evaluate()

For prediction, you can use predict() attribute for predict the data (note: the data should be as same as the input format for the training).

from LogisticRegression import LogisticRegression
from sklearn.metrics import accuracy_score

model = LogisticRegression(lr = 1e-2, lambda_val = 1e-3)
model.load_weights("model.json")
y_pred = model.predict(X)

acc = accuracy_score(y_true, y_pred) 
print(acc)

The config.json file includes the hyperparameters for training, you can customize by yourself in the file

Let me know if are there any issues 😉