一种基于最大Lyapunov指数奇异分解的并行挖掘算法

A parallel mining algorithm based on maximal Lyapunov exponent singular decomposition

针对当前大数据挖掘并行计算采用多元线性回归分析方法导致的计算开销过大、挖掘准确度不高等问题,提出了一种基于最大Lyapunov指数奇异分解的大数据挖掘并行计算方法.该方法对大数据信息流进行高维相空间重构和QR分解,计算大数据流模型的最大Lyapunove指数谱,基于微积分极值理论构建大数据的Lyapunove指数谱的网格分布矩阵,采用奇异值分解方法对参与运算的大数据特征向量矩阵行分解,将大规模的数据挖掘计算问题变为一系列小规模的并行计算问题,实现了大数据挖掘中并行算法的改进.测试结果表明,采用该方法进行大数据挖掘的计算时间较短、内存开销较小、准确度高.

In view of the current data mining parallel computing method of regression analysis leads to excessive computational overhead by using multivariate linear,mining accuracy is not high,in order to improve the efficiency and accuracy of data mining,a large number of maximum Lyapunov Exponent Based on singular decomposition according to the parallel computing method of mining.For phase space reconstruction and QR decomposition of large data flow,the maximum Lyapunov index calculation of large data flow model of the spectrum,constructing the grid distribution matrix data Lyapunov exponent spectrum calculus based on extreme value theory,using singular value decomposition method for participating in the operation of large numbers according to the eigenvector matrix decomposition,large-scale data mining the calculation problem into a series of small scale parallel computing problem,improved algorithm in data mining.Experimental results show that the proposed method is used for large data mining,the computation time is relatively short,the memory cost is small,and the accuracy of data mining has been greatly improved.