This will build an LSTM Autoencoder, train it on a set of normal heartbeats and classify unseen examples as normal or anomalies.
That step for detect abnormal hearbeats:
- Prepare a dataset for Anomaly Detection from Time Series Data
- To build an
LSTM Autoencoder with Pytorch - Train and evaluate your model
- Choose a threshold for anomaly detection
- Classify unseen examples as normal or anomarly
Data source: http://timeseriesclassification.com/description.php?Dataset=ECG5000
The dataset contains 5000 Time Series examples (with ECG) with 140 timesteps. Each sequence corresponds to a single heartbeat from a single patient with congestive heart failure.
An ECG measures and records the electrical activity that passes throught the heart. A doctor can determine if this electrical activity is normal or irregular.
Basically, we have 5 types of heartbeats(classes):
- Normal (N)
- R-on-T Premature Ventricular Contraction (R-on-T PVC)
- Premature Ventricular Contraction (PVC)
- Supra-ventricular Premature or Ectopic Beat (SP or EB)
- Unclassified Beat (UB)
Download data via bash script on Google Colab:
!gdown --id 16MIleqoIr1vYxlGk4GKnGmrsCPuWkkpT
!unzip -qq ECG5000.zip
Need to set up GPU on Google Colab:
device = torch.device("cuda", if torch.cuda.is_available() else "cpu")
!pip install -qq arff2pandas
!pip install -q -U watermark
!pip install -qq -U pandas
%reload_ext watermark %watermark -v -p numpy, pandas, torch, ardd2pandas
CPython 3.6.9 IPython 5.5.0
numpy 1.18.2 pandas 1.0.3 torch 1.4.0 arff2pandas 1.0.1
Using LSTM Autoencoder
The Autoencoder's process is to get some input data, pass it through the model, and obtain a reconstruction of the input. This is a tupe of artificial neural network used to learn efficient data codings is an unsupervised manner.
The aim of autoencoder is to learn a representation (encoding) for a set of data, typically for dimensionality reduction by training the network to ignore signal "noise"
