中文摘要：

我国是一个农业大国,农业受灾严重影响到国家和社会经济的可持续发展。文中以1970-2014年的农业总体、干旱和洪涝受灾面积为原始序列,提出了一种集成NARX神经网络和灰色系统的灾害预测模型。鉴于原始序列随机波动性较大的特点,对其乘以一个序列算子,得到一个相对平滑的新序列;采用灰色系统对新序列进行预测;用NARX网络对序列算子进行预测;将灰色预测值除以NARX网络预测值得到最终的预测结果。通过对农业总体、干旱和洪涝受灾面积进行实证研究,证实了这种预测方法能够有效地提高随机波动性较大序列的预测精度。

英文摘要：

As a big agricultural country, the agricultural disaster severely influences sustainable development of the country. Taking agricultural overall and flood disaster areas as the original sequences, Grey system and NARX network were integrated to predict disasters. They were random fluctuation sequences. New sequences were obtained through multiplying the sequences by nonlinear sequences. The new sequences were relatively smooth. Grey system was used to predict the new sequences and NARX network was utilized to predict the nonlinear sequences. Dividing the grey system predictive value by the NARX network predictive value to form the final forecasting result. Through the empirical research on the agricultural overall, drought and flood, the proposed prediction method can effectively improve the forecasting precision for sequences with large random fluctuation.