Comparing supervised and unsupervised artificial neural network for conducting eco-environment vulnerability assessment

DOI: 10.5176/978-981-08-7466-7_kd-10

Authors: Zhiyuan Yao, Wei Qi, Xiao-ping Pang and Lan Tao

Abstract:

The present study adopts an unsupervised neural network, i.e. the SOM-Ward clustering to conduct eco-environmental vulnerability assessment. The 797 assessment units on Fildes Peninsula are clustered based on the ten assessment indexes; the results effectively display the eco-environment vulnerability classifications on Fildes Peninsula and the factors that impact them. In addition, the results of the SOM-Ward clustering are compared with those predicted by the back propagation (BP) neural network, further verifying the practicability and reliability of the SOM-Ward model. Based on the results of both models, we summarize the characteristics of the eco-environment vulnerability of the Antarctic ice-free areas, providing further materials for the study of Antarctic eco-environmental.

Keywords: Eco-environmental vulnerability assessment; artificial neural network (ANN); Self-organizaingmap (SOM); Ward’s clustering, the back propagation (BP)neural network; GIS data processing;