【摘要】 根据相关系数法筛选出与马尾松毛虫各级危害程度的危害面积相关关系都比较密切的气象因子,分别建立多对多线性回归模型、脉冲响应函数模型、线性差分方程之带输入项的线性自回归模型、人工神经网络模型以及双重筛选逐步回归模型5种马尾松毛虫危害程度分级预报的数学模型。结果表明,运用所建立的5种分级预报模型对预留的4年共16组样本进行预测结果检验,都取得较好的预测效果,其中以线性差分方程之带输入项的线性自回归模型的预测效果最好,人工神经网络模型的预测效果次之,双重筛选逐步回归模型最差。更多还原

【Abstract】 Based on the correlation coefficient method,the meteorological factors with very close correlation with the damage area of Dendrolimus punctatus Walker at all damage degree were selected.5 kinds of mathematical models of the classified prediction of the damage degree of D.punctatus were set up,including multivariate regression model,impulse response function model,linear difference equation model,dual-selective stepwise regression model and artificial neural network model.The results showed that 5 kinds of classified prediction models were used to test the prediction results of 16 samples in 4 years and better prediction effects could be obtained.The prediction effect of linear difference equation model,especially linear auto-regression model with input item was the best,followed by artificial neural network model,and the prediction effect of dual-selective stepwise regression model was the worst.更多还原