Vertical Frequent Pattern Mining from Uncertain Data

In many real-life advanced knowledge-based and intelligent information & engineering applications, data can be uncertain. This leads to the uncertain data mining. In recent years, Apriori-based, tree-based, and hyperlinked array structure based mining algorithms-namely, U-Apriori, UF-growth and CUF-growth, as well as UH-Mine, respectively-have been proposed to mine frequent patterns from probabilistic databases of uncertain data. All these algorithms treat the probabilistic databases "horizontally" as collections of transactions, and each transaction is considered as a set of items associated with existential probability values. In this paper, we consider an alternative representation (i.e., vertical format) of uncertain data such that the probabilistic databases can be viewed "vertically" as collections of items. Each item is associated with a vector that indicates the transactions containing such an item, and each vector entry is associated with an existential probability value. We also propose an advanced knowledge-based algorithm for discovering frequent patterns from this vertical representation of uncertain data.