【摘要】 本文用适当的方法构造“秩和”因子,以便利用更多的因子信息;然后逐步筛选引进“秩和”因子,用训练迭代法建立预报对象不同等级的隶属函数并利用最大隶属原则进行判决和预报,使之具有“学习”过去经验教训的能力和处理摸糊信息的能力,并且适用于任意分布的样本数据,可作趋势及定值预报。跟常用的多级判别法及训练迭代法比较,本文方法适应性更强,效果更好。更多还原

【Abstract】 The "rank sum" factors are properly constructed in order to utilize the information of more factors. By successive introducing "rank sum" factors, the multiple-order discriminate functions are established so that they possess the ability of learning past experiences and treating fuzzy information. They can be applied to the sample data of arbitrary distribution . This method can be used to forcast both the trend and the numerical value of the object to be forecasted. Comparing with the ordinary multiple-order discriminate method and the training iteration method, it is shown that this method can be applied widely to much more situations and often provides a better result.更多还原