门限自回归模型在海洋冰情预测中的应用

APPLICATION OF THRESHOLD AUTO REGRESSIVE MODEL TO PREDICTING MARINE CONDITION

给出了建立门限自回归模型（ＴＡＲ） 的一套实用方法， 用作者提出的改进遗传算法可同时优化门限值和自回归系数，成功地解决了ＴＡＲ建模过程所涉及的大量复杂寻优工作这一难题。实例计算的结果说明了， 这套方法在海洋冰情预测中是可行的和有效的， 在各种自然灾害非线性时序动态预测中具有重要的理论意义和广泛的实用价值。

A practical method is presented for establishing threshold auto regressive (TAR) model. With the improved genetic algorithm by the author, both of the threshold values and auto regressive coefficients can be optimized and the difficult problem of modeling of TAR is solved. The result of calculation example shows that predicting marine ice condition with this method is practical and efficient. As a general way, the scheme has major theoretical valve and wide ranging application for predicting regime of nonlinear time series of different natural disasters.