r2ntab 1.0.3

Interpretable machine learning model for binary classification combining deep learning and rule learning

Developed by M.J. van der Zwart as MSc thesis project (c) 2023

Installation

pip install r2ntab

Preparing data using sample dataset

import torch

from r2ntab import transform_dataset, kfold_dataset

name = 'adult.data'
X, Y, X_headers, Y_headers = transform_dataset(name, method='onehot-compare',
negations=False, labels='binary')
datasets = kfold_dataset(X, Y, shuffle=1)
X_train, X_test, Y_train, Y_test = datasets[0]
train_set = torch.utils.data.TensorDataset(torch.Tensor(X_train.to_numpy()),
torch.Tensor(Y_train))

Creating and training the model

from r2ntab import R2NTab

model = R2NTab(len(X_headers), 10, 1)
model.fit(train_set, epochs=1000)
Y_pred = model.predict(X_test)

Extracting the results

rules = model.extract_rules(X_headers, print_rules=True)
print(f'AUC: {model.score(Y_pred, Y_test, metric="auc")}')
print(f'# Rules: {len(rules)}')
print(f'# Conditions: {sum(map(len, rules))}')

Contact

For any questions or problems, please open an issue here on GitHub.