Abstract: In combination forecast models, the key point is to decide the weight coefficient. This paper does a further study on the estimation of weight coeficient in diferent conditions. On the basis of traditional combination forecast models, we introduce one new intergraded prediction method called the least absolute value which takes the minimum sum of absolute value as the target function. Through the example, we can know that in general, the model of the least absolute value has preferable precision and preferable superiority to traditional models.

Key words: combination forecast model, weight coefficient, the least square method, the least absolute value method

摘要： 组合预测法的关健是确定组合的权重，着重研究了权重确定的不同方法。在讨论传统的组合预测方法的基础上，提出了新的组合预测方法，即以绝对值之和最小为目标函数的最小绝对值法。通过实例，将新方法和传统方法进行对比，验证了新方法的精确性和优越性。

关键词: 组合预测, 权重, 最小二乘法, 最小绝对值法