Deep Learning Based Spatio-temporal Prediction Using LSTM Nerworks

Authors: Wonho Jo*, Seoul National University
Topics: Geographic Information Science and Systems
Keywords: Spatiotemporal analysis, LSTM, Spatial correlation, Temporal correlation
Session Type: Paper
Day: 4/5/2019
Start / End Time: 3:05 PM / 4:45 PM
Room: Roosevelt 3, Marriott, Exhibition Level
Presentation File: No File Uploaded

Spatial phenomena are not determined independently but are affected by temporal and spatial phenomena nearby. In particular, to predict phenomena, it is necessary to construct the model that can deal with temporal correlation as well as spatial one to handle these dependent features. In this study, the machine learning model that explicitly consider the spatio-temporal correlation of spatial phenomena is constructed. In particular, LSTM(long term short-term memory) which has recently enjoyed a great success in the field of deep-learning and proved to be very powerful in analyzing time series data was used for prediction. However, it is challenging problem to extend the LSTM model to the spatial dimension and perform spatio-temporal analysis. This study constructed the LSTM model that considers additional spatial correlation and verified the performance of the model.

Abstract Information

This abstract is already part of a session. View the session here.

To access contact information login