/lstm-fcn

LSTM-FCN applied on ECG dataset

Primary LanguageJupyter NotebookMIT LicenseMIT

lstm-fcn

General presentation:

LSTM-FCN for Long Short Term Memory Fully Convolutional Network is an enhancement of FCN, which was considered as the reference algorithm for time series classification. The LSTM-FCN algorithm achieves better results, compared to other algorithms, for different types of time data such as: weather reports, financial information, psychological signals, or electrocardiograms. For more details about the algorithm: LSTM-FCN

Data set:

We will work on the application of the algorithm to a database containing electrocardiograms of 2 different types, which will allow us to assess the appropriateness of using this new algorithm in a specific case. Link for the dataset: TwoLeadECG

Results (Acuracy):

  • k-NN: 93,06% for k = 1
  • LSTM-FCN: 100,00% !