多元线性回归模型的增量算法

伴随着各领域信息化的发展,数据多呈现出快速、连续流入的特点.面向海量不断更新的数据集,在具有广泛使用价值的线性回归模型中,考虑引入增量算法.通过基于叉积矩阵的增量计算公式,得到最小二乘估计模型的增量算法,并进一步扩展到其他的模型估计量和检验统计量中.该增量算法运用了全部的数据信息,与使用全部数据建模具有完全相同的结果.算法节约了数据读取时间,减小了数据存储传输的压力,从而提高了计算效率.数据仿真实验验证了算法的有效性.

Incremental algorithm of multiple linear regression model

With the development of computer-related technology, people can continuously obtain data faster and faster. Facing with the massive and continuously updated data sets, incremental algorithm was introduced to the popular multiple linear regression analysis. The incremental algorithm of least squares estimation model was derived based on incremental expression of cross product matrix. And further this algorithm was extended to other estimation models and test statistics. The incremental algorithm uses the information of all dataset, which can get the same results with non-incremental methods. This algorithm can save the time in reading and writing data, release the impression on transportation, and thus speed up the computation. Simulation results show that, this algorithm can improve computational efficiency and is very useful in many conditions.