First, you should clone the repo.
git clone https://github.com/King-HAW/ECG_Classification_Baseline
You should install all the dependencies by running:
cd ./ECG_Classification_Baseline
pip install -r requirements.txt
You can run the following command to explore the data distribution.
cd ./get_meta
python get_meta.py
Also, you should running the following command to do the prepeocess. You can get some files after you run the command. trainset.npy and validset.npy are data files. training-nodup.csv and infer.csv are index files.
cd ./preprocess
python preprocessing.py
Just run the train.py, the trainset will be divided into five folds and cross-validation will run automatically. You can change some hyperparameters pre-defined in the script to fit your environment.
python train.py
After the training step is done, run the predict.py to get the 5-fold ensemble result.
python predict.py
You are encouraged to change the code to get better performance. Maybe the self-attention module is a good choice. Have fun and hope you can enjoy it :-)