This repository includes analysis scripts and modules for working with the HCUP readmission dataset.
Place the NRD databases unzipped in the
src folder includes neural models implementation using
PyTorch (version 0.3.x), in addition to training/evaluation workflow and utilities found in the
utilities.py modules respectively.
notebooks folder includes jupyter notebooks to perform the following:
Data processing for the HCUP NRD databases
data_processing.ipynb: Go through the cells of the notebook sequentially to process HCUP readmission dataset.
neural_models_dataset_gen.ipynb: To prepare and generate the dataset in a suitable format for training/evaluating the neural models using 5-folds cross validation.
optimize_hyperparams_neural_models.ipynb: Run this notebook to identify quasi-best hyperparams for neural models.
train_eval_neural_models.ipynb: Train and evaluate the neural models.
train_eval_baseline_models.ipynb: Train and evaluate baseline models.
performance_evaluation_decoded_output.ipynb: To post-process decoded output (i.e. predicted outcomes of the models) for generating performance evaluation.
run_cvAUC_R.ipynb: Run the
cvAUCpackage to analyze AUC (area under the ROC curve) performance of the models on the 5-folds (R kernel should be installed).