Home

Awesome

Codebase for "Estimating Missing Data in Temporal Data Streams Using Multi-Directional Recurrent Neural Networkss (MRNN)"

Authors: Jinsung Yoon, William R. Zame, Mihaela van der Schaar

Paper: Jinsung Yoon, William R. Zame, Mihaela van der Schaar, "Estimating Missing Data in Temporal Data Streams Using Multi-Directional Recurrent Neural Networks," IEEE Transactions on Biomedical Engineering, 2019.

Paper Link: https://ieeexplore.ieee.org/document/8485748

Contact: jsyoon0823@gmail.com

This directory contains implementations of MRNN framework for imputation in time-series data using GOOGLE stocks dataset.

To run the pipeline for training and evaluation on MRNN framwork, simply run python3 -m main_mrnn.py.

Command inputs:

Example command

$ python3 main_mrnn.py --file_name data/google.csv --seq_len 7 
--missing_rate: 0.2 --h_dim 10 --batch_size 128 --iteration 2000
--learning_rate 0.01 --metric_name rmse

Outputs