用车贝雪夫多项式做时间序列预报的一种改进方案

为提高预报准确率，本文对作者以前提出的时间序列预报方法，提出了一种改进方案。这一方案的要点是:1）在计算理想初值的预报时，针对不同情况选择不同的迭代次数；2）在通过迭代终值计算预报时，对不同情况使用不同的拟合误差平均值；3）当拟合误差的符号难以确定时，直接取迭代终值作为预报值。实例计算表明， 新的方案有利于提高预报准确率。

An Improved Scheme for the Forecast of Time Series by Using Chebyshev Polynomials

To raise the accuracy of forecast an improved acheme is presented in this paper for the forecasting method of time series proposed by the author previously. The main points of the scheme are: (1) using different iteFgtion numbers for different initial values of iteration;.(2) using different means of fitting for predicted value obtained from final value of iteration; (3) taking the final value as prediction when it is difficult to difine the sign of final value of iteration. The results indicate that the accuracy of the forecast may be raised by using the improved scheme.