基于最大似然法集成的黄曲条跳甲预警模型

Study on the Early Warning Model of Phyllotreta striolata Based on the Maximum Likelihood Integration

采用最大似然法模型,建立蔬菜黄曲条跳甲的预警模型,并且针对最大似然法一般需要比较多的训练样本才能准确预测的缺点,提出能够显著地提高学习系统的泛化能力的集成算法,即最大似然集成算法以减少对训练样本数量的要求。通过对广东省蔬菜黄曲条跳甲数据验证表明,最大似然集成算法的预警准确率比最近邻算法k、-mean聚类和支持向量机预警准确率都要高。

The early warning model of Phyllotreta striolata in vegetables was set up by using the maximum likelihood model. Aiming at the disadvantage of the maximum likelihood that more training samples were needed for accurate prediction generally, its integration algorithm that could enhance the generalization ability of the learning system significantly was put forward, The maximum likelihood integration algorithm reduced the quantity demands of the training samples. The data of the test on P. striolata in vegetables in Guangdong Province, the accuracy rate of early warning by the maximum likelihood integration algorithm was higher than that by nearest-neighbor algorithm, k-mean clustering and support vector machine.