Code for MLHC2020 paper on glucose prediction using OpenAPS and OhioT1DM data
Usage:
In the terminal run the following commands twice. First with dataset='oaps' and then with dataset='ohio' in preprocess.sh and run.sh files:
chmod +x ./preprocess.sh
./preprocess.sh
chmod +x ./run.sh
./run.sh
In run.sh file, set the following parameters according to the experiment you
are running.
dataset='ohio' (ohio or oaps)
history_window = 12 (number of past glucose values to use. 12 samples means an hour of previous data (frequency = 5 minutes)
prediction_window = 60 (30 minutes or 60 minutes)
dimension = multivariate (univariate or multivariate)
prediction_type = single (single or multi. This refers to single step or multioutput forecasting)
normalize_data = False (True or False. use normalized data or not)
model_name = RNN (['REG','SVR','TREE','ENSEMBLE','LSTM', 'RNN'])
save_results = False (True or False. It will replace old output files in the output directory)
ablation_code = 0 (0 = unfiltered_imputed, 1.filtered_imputed, 2.unfiltered_unimputed 3. filtered_unimputed)