The repository holds the notebooks and code for BGLP Challenge-II at ECAI 2020. The challenge is to predict blood glucose levels of 6 patients provided their CGM (Continous Glucose Monitoring) Levels, Finger Stick values, and other values such as exercise, meal types, etc.
We have experimented with 6 different deep learning models - LSTM, BiLSTM, CNN-LSTM, TCN (Temporal Convolutional Networks), Seq-to-Seq Models, and Transfer Learning. Seq-to-Seq BiLSTM model proved to the best model. The repo provides code (IPython Notebooks) for all the models executed and the code used to convert the provided XML files to CSV files. The research paper corresponding to the repository has been submitted to ECAI.
This work is described in the Proceedings of the 5th International Workshop on Knowledge Discovery in Healthcare Data co-located with 24th European Conference on Artificial Intelligence (ECAI 2020) here: https://ceur-ws.org/Vol-2675/paper22.pdf
If you are using this work, please cite: Bhimireddy, A., Sinha, P., Oluwalade, B., Gichoya, J.W. and Purkayastha, S., 2020. Blood glucose level prediction as time-series modeling using sequence-to-sequence neural networks. CEUR Workshop Proceedings.