一种基于信息嫡的多维流数据噪声检测算法

针对传统数据流聚类算法自适应性不强、对问题的依赖性过高以及聚类质量不够理想、聚类效率低下等缺陷,提出一种基于人工免疫原理的数据流聚类IMStream算法。该算法通过引入衰减函数和时刻权重来反映过去的数据与当前流入的数据在整个数据流中的地位,通过计算抗体期望克隆率E(xi)来限制抗体克隆的数目以及保持抗体的多样性,通过采取网络中的淘汰策略使最终的网络结构更符合原始数据流的内在特性。在真实数据集和人工数据集上的实验表明,IMStream算法比传统的数据流聚类算法具有更好的性能。

Entropy-based Algorithmin for Noise Detection in Multi-dimensional Stream Data

In traditional data stream clustering algorithm,adaptive problem is not strong,the dependence is too high and clustering quality is not ideal,clustering efficiency is low,this paper presented a data stream clustering algorithms based on artificial immune principle-IMStream.This algorithm introduces attenuation function and moment weight to reflect the past data weights with the current data in the data stream of position,through calculating antibody expect cloning rate E(xi) to limit the antibody cloning number and keep the diversity of antibodies,take out eliminated strategy to made the final network structure more to accord with the original data stream inherent characteristics.The experiments based on the real data set and artificial data set show that the algorithm IMStream has better clustering performance.