论文部分内容阅读
采取归一化对10 a内的农田有效灌溉面积进行处理,将处理后的数据作为标准样本,通过BP神经网络与向量机回归构建了农田有效灌溉面积模型。预测结果证实,向量机预测方法的精度更高,其泛化能力有明显提升,根据预测分析,其主要误差仅有BP神经网络总误差数的12.6%,是非常有效的灌溉面积预测方法。基于此,通过分析预测方案的有效性,并且对模型构建的方法与原则进行研究,提出了正确的变化趋势。
The effective irrigation area of farmland within 10 years was treated by normalization. The processed data were used as standard samples, and the effective irrigated area model of farmland was constructed by BP neural network and vector machine regression. The prediction results show that the accuracy of the SVM prediction method is higher and the generalization ability is obviously improved. According to the prediction analysis, the main error is only 12.6% of the total error of the BP neural network, which is a very effective method for forecasting the irrigated area. Based on this, by analyzing the effectiveness of the prediction scheme and studying the methods and principles of the model construction, the correct trend is proposed.