dySTGNN Model Performance (Figure 3) from "dySTGNN: a spatiotemporal graph learning approach for dynamic prediction of acute respiratory failure using high-dimensional ICU Data"
The bar graphs show our dySTGNN model outperforming GRU, LSTM, TCN, and Transformer models across different data resolutions (5, 15, 45 minutes) and time series durations (1, 3, 5, 7 days). Subfigures (a), (b), and (c) illustrate dySTGNN's superior performance for varying resolution lengths through AUC ROC, AUC PR, and F1 Scores, respectively. Subfigures (d), (e), and (f) illustrate dySTGNN's superior performance with increasing time series lengths through AUC ROC, AUC PR, and F1 Scores, respectively. Metrics are from the HiRID-ICU-Benchmark test set, with variability shown by error bars from 10 independent runs.
Refer to the associated publication "dySTGNN: a spatiotemporal graph learning approach for dynamic prediction of acute respiratory failure using high-dimensional ICU Data" for details (under review).
For inquiries regarding the contents of this dataset, please contact the Corresponding Author listed in the README.txt file. Administrative inquiries (e.g., removal requests, trouble downloading, etc.) can be directed to data-management@arizona.edu