基于最小一乘的灰色预测系列模型

Grey Prediction Models Based on Least Absolute Deviation Criteria

针对基于最小二来准则的传统灰色预测模型的参数辨识稳健性较差,甚至会出现病态性问题,提出了基于最小一乘准则的参数辨识方法,并给出了求解该参数的简便算法。最后通过实例说明,与最小二乘准则比较,基于最小一来准则的各类灰色预测模型能够有效降低异常值的干扰,弥补最小二乘法的不足,提高了各类灰色预测模型的适用性。

The parameter identification in classical grey models based on least square regression lacks robustness and often exists ill-conditioned problems. We extend grey prediction models with least absolute deviation criteria and give a simple algorithm of parameter identification. Compared with the least square regression method,the results show that the models based on least absolute deviation regression can reduce the interference of outliers, improve the robustness and exhibit wide adaptability.