摘要
本文对一种新兴统计方法—投影寻踪回归(Projection Pursuit Regression)做了较为详尽的介绍,并提供了一个应用实例。研究结果表明:采用PP回归方法较之以往的多元分析法具有许多突出的优点,诸如可以引用多个预报因子,能较好地解决非正态或非线性问题以及拟合效果与预报效果较为一致等等。 PP回归属于一种非参数统计方法,它注重分析数据本身的內部结构,事先无需假定函数类型,能够自动排除与数据结构、特征无关或关系很小的变量干扰,并能有效地克服“维数祸根”,使我们能够充分利用现有信息,提高干旱长期预报工作的水平。同时由于PP回归方法参数调整灵活、弹性好、适应性强,因此将有广阔的应用前景。
A new developing statistical method, PPR (Projection Pursuit Regression), is introduced in this paper while a practical example is presented.The results of study indicates that the PPR method features some prominent advantages, such as that multiple factors can be introduced, nonnormal and non linear problems can be solved better and fitting result is consistent with prediction result etc. With the use of PPR method the result of data-processing is more satisfactory than that of multiple regression method. The accuracy of fitting and prediction all come up to 75%.PPR is a nonparametric statistical method. It features that flexibalc parameter adjusting, good clesticity and wide adaptability, therefore it has wide application prospect.
