Rock: A robust clustering algorithm for categorical attributes

ROCKET: A Robust Parallel Algorithm for Clustering Large-Scale Transaction Databases

“…ROCK [4] is a clustering algorithm for the objects with categorical attributes. The transaction is a kind of categorical objects.…”

“…The sim() measure can be differently defined according to the target database and is defined as Eq. (1) for two transactions T 1 and T 2 [4]:…”

“…The transaction database is composed of a number of transactions, each of which can have a different number of items. A typical example is a market basket database [4], [6]; a transaction is a set of products simultaneously purchased by a customer at a market. The issue in clustering the transaction database is creating good clusters (or groups) of similar (or highly correlated) transactions and then utilizing them in various applications.…”

“…The previous clustering algorithms could deal with only small-sized databases or could not generate highquality clusters [4], [6], [7]. Moreover, the algorithms were designed without consideration on the up-to-date hardware architecture.…”

“…ROCKET is a divisive hierarchical algorithm using the notion of link in ROCK algorithm [4]. ROCKET is a parallel algorithm making the most of recent hardware architecture.…”