基于正弦余弦算法的异常值线性回归

Linear Regression of Outliers Based on Sine Cosine Algorithm

对给定的数据,文章分别建立了以残差的平方和及绝对值和为目标的线性最小二乘与最小一乘模型,采用正弦余弦算法计算模型参数;然后应用于无异常值的模型和包含异常值的模型,计算结果发现异常值对最小二乘有着较大的影响,而对最小一乘的影响较小;表明最小一乘具有较好的稳健性。

For the given data, this paper establishes linear least squares and least absolute deviation models respectively aiming at the sum of squares and sum of absolute values of residuals, calculating the parameters of models by sine cosine algorithm,and then applies it to models with and without outliers. The calculation results find that the outliers have a great effect on the least squares, but a little on the least absolute deviation, which indicates that the least absolute deviation is of good robustness.