search

liyaguang / DCRNN

Implementation of Diffusion Convolutional Recurrent Neural Network in Tensorflow
MIT License
1.22k stars 400 forks source link

how to deal with the missing data? #18

Closed vision-zhao closed 5 years ago

vision-zhao commented 5 years ago

there are so many missing data and nil data in the METR.h5, so I'm wondering how you deal with it when you are producing your result?

I will be deeply grateful to you!

liyaguang commented 5 years ago

Currently, we just ignore the missing value when computing the training loss. You may also refer to [1][2] for dealing with missing data in time series data. [1] Zhengping Che, Sanjay Purushotham, Kyunghyun Cho, David Sontag, and Yan Liu, Recurrent Neural Networks for Multivariate Time Series with Missing Values, Scientific Reports (SREP), 8(1):6085, 2018. [2] Dingxiong Deng, Cyrus Shahabi, Ugur Demiryurek, Linhong Zhu, Rose Yu and Yan Liu, Latent Space Model for Road Networks to Predict Time-Varying Traffic, International Conference on Knowledge Discovery and Data Mining (KDD), 2016.

RobinLu1209 commented 5 years ago

Hi, I still have some problems about missing data. It's a good choice to ignore the missing value when computing the training loss, but what's your action to train? There will be data mutation and abnormality in training.