CAJ | 학술논문

　　为了利用气象因子预测松毛虫的发生面积，基于辽宁省阜新县1983—2008年气象资料与松毛虫发生面积等资料进行相关分析，将筛选出的气象因子作为预报因子，通过多元回归法和人工神经网络法对松毛虫发生面积进行模拟和预测。结果表明，与松毛虫发生面积显著相关的5个气象因子包括：上一年12月平均最低温度、上一年11月平均相对湿度、上一年9月降水量、本年2月降水量和本年3月降水量；人工神经网络法的模拟和预测精度均优于多元回归法，多元回归法的预测精度58.2%，人工神经网络法的预测精度为83.6%；人工神经网络法更适用于辽宁地区松毛虫发生面积的预报。
　　위료이용기상인자예측송모충적발생면적，기우요녕성부신현1983—2008년기상자료여송모충발생면적등자료진행상관분석，장사선출적기상인자작위예보인자，통과다원회귀법화인공신경망락법대송모충발생면적진행모의화예측。결과표명，여송모충발생면적현저상관적5개기상인자포괄：상일년12월평균최저온도、상일년11월평균상대습도、상일년9월강수량、본년2월강수량화본년3월강수량；인공신경망락법적모의화예측정도균우우다원회귀법，다원회귀법적예측정도58.2%，인공신경망락법적예측정도위83.6%；인공신경망락법경괄용우료녕지구송모충발생면적적예보。
In order to forecast the occurrence area of Dendrolimus by using meteorological factors, the author analyzed the relationships between meteorological factors and occurrence area of Dendrolimus with the data from 1983 to 2008 in Fuxin County, Liaoning. The occurrence area of Dendrolimus was simulated and predict on the selected meteorological factors as forecast factors, using multiple element regression and artificial neural network methods. The results showed that: 5 meteorological factors were significantly correlated with the occurrence area, including the mean minimum temperature of preceding December, the mean relative humidity of preceding November, the precipitation of preceding September, the precipitation of current February and current March. The simulation and prediction accuracy rate of the artificial neural network method was better than that of the multiple element regression method, the multiple element regression method reached over 58.2% while the artificial neural network method reached over 83.6%. The artificial neural network method was more appropriate for the occurrence area forecast of Dendrolimus in Liaoning.