Welcome to wrapped-ml

wrapped-ml is a machine learning library that wraps Keras, Scikit-learn and XGBoost into a common framework.

NB this package is in development elsewhere and updates will be posted to this repository sparingly, as they finish. In the meantime, feel free to use what's here!

---

Usage

Some brief example usage for a classifier:

# Import type of model you want to use e.g. ClassifierNNClass, ClassifierSklearnClass or custom (as here).
from ExampleNNClassifierClass import ExampleNNClassifierClass

# We will build NN Classifier, let's specify the architecture as a list of layer names and their params
network_architecture: list = [('Dense', {'units': 8, 'activation': 'relu'}),
                              ('Dropout', {'rate': 0.1}),
                              ('Dense', {'units': 1, 'activation': 'sigmoid'})]
# Build params that are passed to the base model (Note, you'll also have to specify input shape)
network_input_data: dict = {'network_architecture': network_architecture}
# This example class takes a string as a dummy example custom input
dummy_input: str = "test"

# Build the classifier
my_classifier = ExampleNNClassifierClass(dummy_input, network_input_data)

# Build and train the model using a test and train Dataframe
my_classifier.train_model(train_df=some_train_df, test_df=some_test_df)

# Optionally calibrate probabilties (dont use training data!)
my_classifier.calibrate_probabilities(x_data, y_data)

# Evaluate on some validation data, with orptional parameters in Dict: testing_func_args
my_classifier.evaluate_on_test_data(validation_input, validation_output, testing_args=testing_func_args)

# Make a prediction
predictions = my_classifier.predict(new_observations)

Dependencies

• Python 3.x
• dill~=0.3.3
• joblib~=1.0.1
• matplotlib~=3.4.2
• numpy~=1.19.5
• scikit-learn~=0.24.2
• tensorflow~=2.5.0
• xgboost~=1.4.2