基于偏最小二乘回归的黑龙江省大豆生产安全预警分析

为消除先行警兆指标之间的多重相关性,引入偏最小二乘回归方法,建立黑龙江省大豆生产警情指标与滞后若干年的先行警兆指标的模型,利用警兆指标的先行值得到警情指标的未来值,并确定警限进行预警分析。结果表明,与一般的多元回归相比,偏最小二乘回归方法避免了自变量之间多重相关性带来的问题,能够得到与实际经验相符合的模型,保证了预警结果的准确性。

Early Safe Warning Analysis of Soybean Production in Heilongjiang Province Based on Partial Least Square Method

In order to avoid the multiple correlativity among the advance alarm aura indexes,the partial least square method was introduced to build the warning index and the advance alarm aura model which lagged behind others by several years for the soybean production in Heilongjiang Province.The future value of warning condition was obtained by the advance value of alarm aura index,the limits of warning were determined and an analysis was made on the early warning.The result showed that,compared with the general multi-dimensional regression law,partial least square method avoids the problems brought by the multiple correlativity between independent variables,and can get the model accompanying with the actual experiences,thus it can ensure the accuracy of the early warning result.