Spatio-temporal Model Based on Back Propagation Neural Network for Regional Data in GIS

Authors:
Jing Zhu;Xiang Li;Lin Du
Affiliations:
Institute of Computer Science, China University of Geosciences, Wuhan, China;Institute of Computer Science, China University of Geosciences, Wuhan, China;Institute of Computer Science, China University of Geosciences, Wuhan, China
Venue:
ISICA '09 Proceedings of the 4th International Symposium on Advances in Computation and Intelligence
Year:
2009

Citing 4
Cited 0

2006 Special issue: Temporal neural networks for downscaling climate variability and extremes

Neural Networks - 2006 special issue: Earth sciences and environmental applications of computational intelligence
Advanced Dynamic-system Simulation: Model-replication Techniques and Monte Carlo Simulation

Advanced Dynamic-system Simulation: Model-replication Techniques and Monte Carlo Simulation
Methods of predicting milk yield in dairy cows-Predictive capabilities of Wood's lactation curve and artificial neural networks (ANNs)

Computers and Electronics in Agriculture
Adaptive fuzzy modeling versus artificial neural networks

Environmental Modelling & Software

Quantified Score

Hi-index	0.00

Visualization

Abstract

This paper focuses on spatio-temporal non-linear intelligent prediction modeling for regional data, and discusses the application of Back-Propagation neural network (BPN) into analysis of regional data in geographic information system (GIS). With their characteristics of space-dependence and space volatility, the regional data determine the accuracy of the prediction model.With consideration of the sectional instability of the spatial pattern, the paper brings forward a modeling method based on regional neural network. First, the space units of researching regions are divided into different sub-regions by improved K-means algorithm based on spatial adjacency relationship, corresponding with sub-regions. Then, a modular BP network is built up, which is composed with main network, gate-network and sub-network. This network is thus named as regional spatio-temporal neural network (RSTN) model. Afterwards, the sub-networks are traiend respectively for every sub-region, and the output of sub-networks is input of main network with adjustment of gate-network The output of main network is predictive results. The spatio-temporal predictive capability of model is measured by average variance rate (AVR) and dynamic similar rate (DSR). At last, the RSTN model and the global BPN model are compared by the analysis of an example: prediction for influenza cases of 94 countries in France. The comparison declares that RSTN model has more powerful prediction capability.