| 离差最大化结合BP神经网络评价烟叶化学品质 |
| |
| 引用本文： | 张勇刚,任志广,徐志强,刘建国,张晓兵,刘化冰,夏琛,程昌合.离差最大化结合BP神经网络评价烟叶化学品质[J].作物杂志,2023,39(1):190-997. |
| |
| 作者姓名： | 张勇刚 任志广 徐志强 刘建国 张晓兵 刘化冰 夏琛 程昌合 |
| |
| 作者单位： | 浙江中烟工业有限责任公司，310024，浙江杭州 |
| |
| 基金项目： | 中国烟草总公司科技重大专项项目(中烟办[2019]69号);浙江中烟工业有限责任公司科技项目(ZJZY2016B002) |
| |
| 摘 要： | 为解决烟叶化学品质现有组合评价方法的不足，探讨离差最大化组合评价法与BP神经网络相结合进行评价。选取4种典型单一评价法，首先采用改进熵权法和AHP法确定指标权重，然后根据离差最大化原理计算组合评价值，最后利用BP神经网络对组合评价值进行反演。结果表明，离差最大化组合评价值与单一评价法相关性较其他组合评价法更高，平均相关系数为0.9822;BP神经网络对组合评价值有较高的预测准确性与稳定性，预测值与实际值相对误差不超过3%，决定系数大于0.9900。说明，离差最大化组合评价法对单一评价法的组合效果更好，BP神经网络提高了组合评价的便捷性。
|
| 关 键 词： | 烤烟 化学品质 改进熵权法 离差最大化 BP神经网络 |
| 收稿时间： | 2021-09-01 |
Chemical Quality Evaluation of Flue-Cured Tobacco Based on Maximization of Deviation and BP Neural Network |
| |
| Zhang Yonggang,Ren Zhiguang,Xu Zhiqiang,Liu Jianguo,Zhang Xiaobing,Liu Huabing,Xia Chen,Cheng Changhe.Chemical Quality Evaluation of Flue-Cured Tobacco Based on Maximization of Deviation and BP Neural Network[J].Crops,2023,39(1):190-997. |
| |
| Authors: | Zhang Yonggang Ren Zhiguang Xu Zhiqiang Liu Jianguo Zhang Xiaobing Liu Huabing Xia Chen Cheng Changhe |
| |
| Institution: | China Tobacco Zhejiang Industrial Co., Ltd., Hangzhou 310024, Zhejiang, China |
| |
| Abstract: | Maximizing dispersion and the BP neural network were proposed as potential combined evaluation methods to address the inadequacies of the present combined evaluation methods of tobacco chemical quality. Four typical single evaluation approaches were chosen for the airticle. First, index weights were determined using the enhanced entropy weight method and the AHP method. Next, the combined evaluation value was calculated using the maximum dispersion principle. Finally, the combined evaluation value was inverted using BP neural network. The results showed that the average correlation coefficient between the maximum deviation combination evaluation value and the single evaluation method was 0.9822, which was higher than that of other combination evaluation methods, and the BP neural network had higher prediction accuracy and stability for the combined evaluation value, the relative error between the predicted value and the actual value was less than 3%, and the coefficient of determination was more than 0.9900. It showed that the deviation maximization combination evaluation method had better combination effects on the single evaluation method, and BP neural network improved the convenience of combination evaluation. |
| |
| Keywords: | Flue-cured tobacco Chemical quality Improved entropy weight method Maximum deviation BP neural network |
|
| 点击此处可从《作物杂志》浏览原始摘要信息 |
| 点击此处可从《作物杂志》下载免费的PDF全文 |
|
