基于双曲方程特征分解的水生态数据挖掘

Data Mining of Water Ecological Based on Feature Decomposition of Hyperbolic Equations

提出一种基于非线性双曲方程广义特征分解的离群数据挖掘方法,并应用于大型分布式数据库离群数据挖掘。采用广义特征分解的方法求解数据集的离群因子,求解非线性双曲方程广义特征Taylor展开。对离群数据的线性部分进行隐格式逼近,对非线性部分进行显格式逼近,挖掘到的离群数据深度和方位信息。构建水生态环境并进行仿真实验。实验结果表明,该算法在对大型分布式数据库的离群数据挖掘中,能很好地挖掘到离群数据张成子空间谱信息,具有较好的数据特征挖掘性能。

The generalized characteristic decomposition method of nonlinear hyperbolic equations is researched in this paper, and it can be applied in data mining of large-scale distributed database. Traditional outlier data mining using the method of hyperbolic equations separate operator alternate difference calculation. In the solving process of second-order hyperbolic equations, the use of non-linear differential operator does not reasonably reflect the nonlinear characteristics of the data, differential operation in parallel is performed in the non - singular matrix, leading to failure mining of some outlier feature points. A kind of outlier data mining method is proposed based on the generalized characteristic decomposition of nonlinear hyperbolic equations. Using generalized characteristic decomposition method solve the outlier factor of data set. Nonlinear hyperbolic equations is solved to expand the generalized characteristics of Taylor. The line- ar part of outlier data is conducted implicit scheme approach, and the nonlinear part is made explicit scheme approach, to mine the depth and orientation information of outlier data. Construction of the water environment, the simulation experiments show that the algo- rithm for outlier data mining of large-scale distributed database can well mine the Zhang Chengzi spatial spectrum information of outlier data, with better performance of the data characteristics mining.