中文摘要：

提出迭代式分割与合并的算法（IDMSVD），以改善最小平方估计的奇异值分解法在估计参数时非常耗费时间以及内存空间的问题。基于此又提出一种使用云计算Hadoop平台MapReduce实现的算法，称为分布式IDMSVD算法。实验结果显示，IDMSVD可以有效地改善SVD求最小平方解耗费运行时间与内存空间的问题，且分布式IDMSVD算法可进一步改善IDMSVD的运行时间。

英文摘要：

Iterative divide-and-merge SVD-based least squares estimator （IDMSVD） is proposed to improve time-consuming and memory space in estimating parameters which come from the least squares estimation based on singular value decomposition method. Moreover, the algorithm which uses cloud computing platform Hadoop MapReduce is proposed, called distributed IDMSVD algorithms. Experimental results show that IDMSVD can effectively improve the SVD in running time and memory space consuming, and distributed algorithm can be further improved IDMSVD in running time.