基于双警戒参数模糊自适应谐振网络的数据流聚类

Data Stream Clustering Based in Dual Vigilance Fuzzy Adaptive Resonance Network

基于一般自适应谐振网络(ART)的聚类受限于其内部的明确表示。通过在单个网络中加入更严格(数据压缩)和更宽松(聚类相似度)的两个警戒参数,扩展模糊ART(Fuzzy ART)的能力,研究一种新的无监督神经网络用于流数据稳定在线聚类,即双警戒参数模糊自适应谐振网络(DVFA),DVFA提高捕获任意形状聚簇的能力,且在数据流聚类方面的性能超过模糊自适应谐振网络。

Clusters by generic Adaptive Resonance Theory(ART)networks are limited to their internal categorical representation.This paper extends the capabilities of Fuzzy ART by incorporating multiple vigilance thresholds in a single network,which are stricter(data compression)and looser(cluster similarity)vigilance values,to study a new unsupervised neural network for stable clustering on stream data.It is Dual Vigilance Fuzzy ART(DVFA),which improve the ability to capture clusters with arbitrary geometry.DVFA can outperform Fuzzy ART while yielding a statistically-comparable performance in data stream clustering.