数据流中随机型分形维数计算方法研究

Research of Stochastic Fractal Dimension Calculation Algorithm in Data Stream

分形维数能够有效地描述数据集,反映复杂数据集中隐含的规律性,基于分形理论的数据挖掘算法通常都涉及到分形维数的计算。但是现有的分形维数计算方法的时间复杂度和空间复杂度都比较高,大大降低了算法的效率,使算法很难适应高速、海量的数据流环境。因此,总结分析了现有的几种分形维数计算方法,并提出一种随机型方法,利用固定的内存空间快速估计数据流的关联维数。最后通过与现有算法进行对比实验,证明了这一随机型算法的有效性。

Fractal dimension can describe the data set effectively and can reflect the hidden regularity of the complex data set.Data mining algorithms based on fractal theory are usually related to the calculation of fractal dimension.But most of the existing fractal dimension calculation algorithms are with high time complexity and space complexity,which greatly reduces the efficiency and is not applicable for data stream with high-speed and massive data.In this paper,se-veral existing fractal dimension calculation algorithms were analyzed and a stochastic fractal dimension calculation algorithm were proposed to fast estimate the correlation dimension in fixed space.The comparative experiment and analysis demonstrate the effectiveness of this stochastic fractal dimension calculation algorithm.