Title : Prediction of ECG Abnormality Based on Recursive Neural Network
This project is based on the mxnet framework
Directory:
---ECG5000
---mode
---train_vae.py
---train_lstm.py
---train_vae_lstm.py
---README.md
ECG5000:
1.A folder where data is stored
2.Data is introduced : a 20-hour long ECG recorded from a 48- year-old male with severe congestive heart failure. This record has 17,998,834 data points containing 92,584 heartbeats. The data description address is[data description][1]
mode:
---lstm.py
---vae.py
---vae_lstm.py
Train mode function file:
train_lstm.py contains functions to train LSTM mode
train_vae.py contains functions to train VAE mode
train_vae_lstm contains functions to train VAE_LSTM mode
introduce vae_lstm:
VAE_LSTM is to input the hidden variable Z generated by the VAE coding layer after training into the modified LSTM model to mine the characteristics of the whole data and time series
Test accuracy:
LSTM : 95.2953%
VAE_LSTM : 98.4985%
[1]: http://timeseriesclassification.com/description.php?Dataset=ECG5000# ECG