Statistical Estimation for Single Linkage Hierarchical Clustering

摘要: Clustering is a key component of most detectors cyber-attacks , and increasingly, for both theoretical practical reasons, methods that produce hierarchical clusterings (dendrograms) are being deployed in this context. In particular Single Linkage Hierarchical (SLHC) attracting con­ siderable interest. Existing clustering algorithms take no account uncertainties the data. paper, we derive statistical model estimation dendrograms, taking into uncertainty (through noise or corruption) distances among data points. We focus on just hierarchy partitions afforded by dendrogram, rather than heights latter. The concept estimating "dendrogram structure" under SLHC introduced, an approximate maximum likelihood estimator (MLE) dendrogram structure described. proposed demonstrated simple Monte Carlo simulation which demonstrates MLE method performs better obtaining correct structure. I. INTRODUCTION Fast detection critical aspect cyber- security, many implemented potential available purpose. underpinning technology attack algorithms; these play important role where learning analysis have to be performed unsupervised fashion. Specifically,