| 摘要: |
| 【目的】小麦产量估测为有关部门制定政策和经济计划提供依据，在粮食宏观调控中发挥重要作用。【方法】文章采用支持向量回归方法估测冬小麦产量，以陕西省关中平原的5个市（西安市、宝鸡市、铜川市、渭南市、咸阳市）作为研究区，将2011—2016 年研究 区内冬小麦4个生育时期（返青期、拔节期、抽穗—灌浆期、乳熟期）的条件植被温度指 数、叶面积指数和每年的单产数据作为总样本，划分训练集和试验集。基于MATLAB 平台 和LIBSVM3.23 软件包，建立研究区冬小麦产量回归预测模型，得到产量预测结果并评价模 型精度。【结果】回归模型的决定系数为0.88，平均绝对百分比误差为6.12%，均方根误差 336.39 kg/hm2。【结论】支持向量回归模型拟合较为理想，有较高的预测精度和较强的泛化能力。回归时的重要参数有惩罚因子C和核参数σ，其中核参数σ对模型精度影响更大。研究表明用该回归模型进行冬小麦产量预测是可行的，支持向量回归方法在粮食产量预测领域 有良好的应用前景。 |
| 关键词: 冬小麦产量 支持向量回归 估产 |
| DOI：10.12105/j.issn.1672-0423.20190602 |
| 分类号: |
| 基金项目:国家自然科学基金重点项目“基于“三位一体”空间抽样理论研究及其二联查找表研建”（41531179） |
|
| Study on yield prediction of winter wheat in Guanzhong Plain based on SVR |
|
Zeng Yan1, Wang Di※1, Zhao Xiaojuan2
|
|
1.Institute of Agricultural Resources and Regional Planning，Chinese Academy of Agricultural Sciences/Key Laboratory of Agricultural Remote Sensing，Ministry of Agricultural and Rural Affairs Beijing 100081，China;2.Center of agriculture and husbandry remote sensing of Qinghai province，Xining 810008，China
|
| Abstract: |
| ［Purpose］Prediction of wheat yield can provide basis for relevant departments to formulate policies and economic plans，and play an important role in macro-control of grain. ［Method］Support Vector Regression（SVR）method is used to estimate winter wheat yield. Five cities in Guanzhong Plain of Shaanxi Province（Xi’an City，Baoji City，Tongchuan City， Weinan City and Xianyang City）were selected as the study area. Vegetation Temperature Condition Index（VTCI）and Leaf Area Index（LAI）of four growth stages（green-turning stage，jointing stage，heading-filling stage and milk-ripening stage）of winter wheat in the study area from 2011 to 2016 were selected. VTCI，LAI and annual yield data were used as total samples which were divided into training set and experimental set. Based on the MATLAB platform and LIBSVM3.23 software package，the study established the prediction model of winter wheat yield in the study area，got the output prediction results and evaluated the accuracy of the model.［Result］The determinant coefficient of the regression model was 0.88. The average absolute percentage error was 6.12%. The root mean square error of the model was 336.39 kg/hm2.［Conclusion］The support vector regression model fitting has high prediction accuracy and strong generalization ability. The important parameters in regression are penalty factor C and kernel parameter σ，in which kernel parameter σ has a greater impact on accuracy of model. The results show that it is feasible to use this regression model to predict winter wheat yield，and the support vector regression method has a good application prospect in the field of grain yield prediction. |
| Key words: winter wheat yield support vector regression yield estimation |
