The MTML-NN.py file is for the doctor/nurses job-role based multitask multilabel model proposed in the paper:
Han Yu, Akane Sano, "Forecasting Health and Wellbeing for Shift Workers Using Job-role Based Deep Neural Network", EAI MobiHealth 2020 - 9th EAI International Conference on Wireless Mobile Communication and Healthcare
Enviornment:
- Python version: 3.6
- Python libraries: --numpy, --pandas, --sklearn, --tensorflow-gpu: 1.12.0 --keras, --pickle
This implementation includes the models for 3 different tasks -- binary classification, 3-class classification, and regression. To use MTML_NN model, you can import our MTML_NN model and initialize it by for example:
model = MTML_NN(featureDimension = X_train.shape[1], task_type = 'reg')
Note that the last column if the features input should be the cohorts of samples, e.g. nurses(N) or doctors(D).
Then, you can train and validate the model by
model.train(X_train, y_train)
model.predict(X_test)