How phantomcoder9632/apneoa-event-prediction-using-ppg-signal works
This repository contains scripts and saved models for PPG-based sleep-apnea forecasting and analysis. The intended task is to use a 30-second PPG window to predict apnea 5 seconds ahead; the available training and evaluation scripts also describe patient-level splits and five-fold validation. The tree does not contain the dataset or several scripts named in the README, so those broader pipeline stages cannot be mapped. The excerpts show data and model loading, but do not establish all artifact handoffs.
Main entry points: testing_final.py Interactive Evaluation · prediction_window_plot.py Prediction Examples · training.py Train Predictor · preapnea_vs_baseline.py Pre-Apnea Analysis
Components 8
Forecast Model
- Train Predictor
training.py - PyTorch Weights
my_final_apnea_predictor.pth
Evaluation
- Interactive Evaluation
testing_final.py - Keras Weights
my_final_apnea_predictor.keras - Prediction Examples
prediction_window_plot.py
Signal Analysis
- Pre-Apnea Analysis
preapnea_vs_baseline.py
Result Visualization
- Metric Plots
visualization_plots.py
Other components
- PPG Dataset
Connections 6
- Train Predictor → PPG Dataset: loads
training.py - Interactive Evaluation → PPG Dataset: loads
testing_final.py - Interactive Evaluation → PyTorch Weights: loads
testing_final.py - Prediction Examples → PPG Dataset: loads
prediction_window_plot.py - Prediction Examples → Keras Weights: loads
prediction_window_plot.py - Pre-Apnea Analysis → PPG Dataset: loads
preapnea_vs_baseline.py