Skip to content

We propose to continuously predict the whole $station \times feature$ map in the next 1 to 48 hours with a convolutional LSTM (ConvLSTM, long short-term memory), which replaces the fully connected network in the classical LSTM with convolutional operations.

Notifications You must be signed in to change notification settings

bruce803/Air-quality-prediction

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

30 Commits
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Air-quality-prediction

We propose to continuously predict the whole $station \times feature$ map in the next 1 to 48 hours with a non-stationary convolutional LSTM ( long short-term memory), which replaces the fully connected network in the classical LSTM with detrending and convolutional operations.

Prediction on consecutive four frames

Visualization of the one $station \times feature$ frame predicted by the seq2seq model (left) vs the ground truth (right).

Visualization of the one $station \times feature$ frame predicted by the seq2seq model (left) vs the ground truth (right).

Visualization of the one $station \times feature$ frame predicted by the seq2seq model (left) vs the ground truth (right).

Visualization of the one $station \times feature$ frame predicted by the seq2seq model (left) vs the ground truth (right).

Baseline

Prediction for seven cities of China, the Beijing, Shanghai, Chengdu, Wwuhan, Gguangzhou, Xi'an, and Nanjing.

Visualization of the one $station \times feature$ frame predicted by the seq2seq model (left) vs the ground truth (right).

About

We propose to continuously predict the whole $station \times feature$ map in the next 1 to 48 hours with a convolutional LSTM (ConvLSTM, long short-term memory), which replaces the fully connected network in the classical LSTM with convolutional operations.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published