摘要
本文提出了强迫“主要变量”的办法,解决了预报方程中自变量的最优子集问题,并以实例证明:预报的准确率有了很大的提高。
In this paper, by classfying auto variables to one of functions that we considered X, Xa, lnX, eax and obligated imputing 'important variables' to regression equation , we offered the optimal prediction and discriminant method when auto variables may not be linear. The advantages of the proposed method are also discussed by two examples.
