| 基于GA的果树仿形喷雾神经网络混合模型研究 |
| |
| 引用本文： | 林惠强,刘财兴,洪添胜,肖磊,高稳猛.基于GA的果树仿形喷雾神经网络混合模型研究[J].农业工程学报,2007,23(10):167-171. |
| |
| 作者姓名： | 林惠强 刘财兴 洪添胜 肖磊 高稳猛 |
| |
| 作者单位： | 华南农业大学信息学院,广州,510642 |
| |
| 基金项目： | 国家高技术研究发展计划(863计划);国家自然科学基金 |
| |
| 摘 要： | 基于BP神经网络研究果树施药仿形喷雾参数之间的关系发现，BP神经网络无法避免不稳定性和局部极小的局限性，为此，该文利用遗传算法优化BP神经网络的权系数，构建遗传算法和BP的果树仿形喷雾神经网络混合模型。研究结果表明：混合模型的计算精度比BP模型要高，其平均相对误差由0.05减为0.019，均方误差由0.005减为0.002。在预测应用中，不仅绝对误差减少，而且合格率从原来的60%提高到80%，较好地解决了单纯神经网络模型的不稳定性，避免局部极小的缺点。
|
| 关 键 词： | 仿形喷雾 神经网络 遗传算法 模型 果树 |
| 文章编号： | 1002-6819（2007）10-0167-05 |
| 收稿时间： | 2006/10/27 0:00:00 |
| 修稿时间： | 2006-10-27 |
Neural network mixed model for profile modeling spray of fruit trees based on GA |
| |
| Lin Huiqiang,Liu Caixing,Hong Tiansheng,Xiao Lei and Gao Wenmeng.Neural network mixed model for profile modeling spray of fruit trees based on GA[J].Transactions of the Chinese Society of Agricultural Engineering,2007,23(10):167-171. |
| |
| Authors: | Lin Huiqiang Liu Caixing Hong Tiansheng Xiao Lei and Gao Wenmeng |
| |
| Institution: | College of Information, South China Agricultural University, Guangzhou 510642, China;College of Information, South China Agricultural University, Guangzhou 510642, China;College of Information, South China Agricultural University, Guangzhou 510642, China;College of Information, South China Agricultural University, Guangzhou 510642, China;College of Information, South China Agricultural University, Guangzhou 510642, China |
| |
| Abstract: | The relationship among the parameters for profile modeling spray of fruit trees based on BP neural network shows that BP neural network cannot avoid instability and local infinitesimal. In order to overcome the localization of BP neural network, in which can not avoid instability and local infinitesimal, GA was used to optimize the weight coefficient, and a model combining BP with GA was set up. The result shows that, the accuracy of the mixed model is higher than the BP model, the average relative error falls down from 0.05 to 0.019, and mean square error down from 0.005 to 0.002; during the forecast, the relative error cuts down, and the eligible rate boosts from 60% to 80%. The mixed model can solve the instability which the simple model has, and avoid the disadvantage of local infinitesimal. |
| |
| Keywords: | profile modeling spray neural network Genetic Algorithm(GA) model fruit tree |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ |
| 点击此处可从《农业工程学报》浏览原始摘要信息 |
| 点击此处可从《农业工程学报》下载免费的PDF全文 |
|
