基于随机权神经网络的农业旱灾脆弱性评价

Vulnerability evaluation and analysis for agricultural drought based on neural network with random weights

农业旱灾脆弱性综合评价是增强农业旱灾风险管理水平的重要方法。运用随机权神经网络法对四川省2000—2015年农业旱灾脆弱性水平进行综合评价分析,检验随机权神经网络评价模型的性能并将其与RBF神经网络和传统的BP网络模型进行比较。研究结论表明:四川省农业旱灾脆弱性水平总体上呈减弱趋势,经济的快速增长是其主要驱动因素;随机权神经网络评价模型的各项性能均优于RBF和BP神经网络,该研究结论为农业旱灾风险管理提供了新思路和新方法。

Comprehensive evaluation of agricultural drought vulnerability is an important method to enhance agricultural drought risk management. The level of agricultural drought vulnerability in Sichuan Province from 2001 to 2015 was assessed by NNRW network method, and the performances of NNRW evaluation model were tested. The study results indicate that, the degree of agricultural drought in Sichuan shows a weakening trend. The rapid economic development is the major driving factor. In conclusion, the various properties of NNRW network method are advantageous over the RBF and BP network model due to result of comparison. This study presents a new idea and method for agricultural drought risk management.