一种基于支持向量回归的梨黑星病预测方法

SVM-Based Method of Pre-Decison on Pear Scab

当前果树病害预测方法,存在适应性差、预测结果拟合度较低的问题.提出一种基于支持向量回归的预测方法框架SVR-D1.0,该方法利用核校准进行核心函数的选择,具有动态更新模型的特点.将黄河故道地区砀山酥梨黑星病为例进行测试的数据,与现有方法以及实测数据进行相关性统计分析.实验表明,在对砀山酥梨的黑星病预测上,该方法与现有方法相比,在实效性、拟合度和准确率上具有较为显著的优势.该方法不仅简便可行,而且可以周期性更新预测模型,具有一定的普适性.

At present, there are poor adaptability and poor effect of forecasting in the method of Pre - Decison on fruit diseases. In this paper, a forecast method SVR-D1.0 based on support vector regression is put forward, the method can select kernel function using kernel alignment, can select keen correlative features repeatedly and update model dynamically. Relativity statistical analysis was conducted between the real data and the forecasting data of Dangshansu pear scab, which show the method is more super and more valid than the current method in efficiency and precision of forecast the occurrence tendency of Dangshansu pear scab. The method is simple and feasible, but also has certain universality.