最优化因子处理及加权多重回归模型
THE OPTIMAL TRANSFORMATION OF PREDICTOR AND WEIGHTED MODEL OF MULTI-REGRESSION
-
摘要: 因子的优劣是回归分析的关键.故在建立回归方程时,一般除对预报因子进行直线相关普查外,还要进行非线性相关普查.如将原因子x用x-1、x1/2、x2、ex、Inx等函数形式进行变换.但由于这些函数形式有限,故不一定能找到最优的表达形式;为此冯耀煌等仅给出了xa和eax两种通式,其中a为待定参数,由最优化方法求得,从而大大提高了因子的相关系数.Abstract: By discussing the relationship between predictor and predictand (i.e.yield) in yield prediction,the basic form of linear and nonlinear transformation of predictor was summed up.The calculation method of parameter in the form was given out,based on the principles of optimization method,and at the same time the optimal predictors were selected.In addtion,the weighted model of multiregression for yield prediction was established,based on the stage of plant.