Abstract: This paper puts forwards the least absolute value method，discusses the characteristic and the course of solving of the least absolute value. And compares least absolute value method and the traditional method-the least two multiplication. Through the example，we can know that at the definite conditions，the model of the least absolute value has preferable precision of forecasting and preferable value of using.

Key words: linear model, distinguish, equivalence, the point of key

摘要： 提出用“最小绝对值法”辨识线性模型，讨论了最小绝对值法解的特性及使用其辨识线性模型的求解过程。并以算例将最小绝对值法与传统的最小二乘法在模型辨识方面进行了比较，在一定条件下使用最小绝对值法辨识的线性模型的预测精度要比使用最小二乘法得到的模型的预测精度高，故而在一定范围内有良好的使用价值。

关键词: 线性模型, 辨识, 等价性, 关键点