Navigation Rules for Exploring Large Multidimensional Data Cubes

Abstract

Navigating through multidimensional data cubes is a nontrivial task. Although On-Line Analytical Processing (OLAP)

Measuring interestingness of discovered skewed patterns in data cubes

“…Aggregation often hides characteristics of the detailed data: an "extremely high" value and an "extremely low" value can be aggregated to a "moderate" value, hiding both extreme values. Using the lowest level of granularity, the problem of hiding a surprise as a side effect of aggregation is avoided [17]. Fig.…”

“…Here α and ffiffiffiffiffi b 1 p are the level of significance and skewness of the random variable fact [3,17]. For instance, if the profit for a node "category = drinks" is positively skewed at α = 0.05 and ffiffiffiffiffi b 1 p ¼ 2:63, the corresponding sk-navigation rule would be "product category = drinks → profit = sk-high [0.05, 2.63]."…”

“…Also, the discovery of interesting summaries in generalization space graphs has been studied in [17,22], using the expectations of users, and relative variance as an interestingness measure. However, unlike the method described in [17,22] we do not require users to specify probability distributions. In addition we deal with the lowest level transactions instead of summaries.…”

“…Another recent approach is presented in [5] where the authors use Simpson's paradox as the basis for defining surprises in multidimensional data but do not address the navigational techniques to reach surprises of interest. Recently, in [17], skewness based navigation rules for data cube exploration were proposed. These rules provide navigation support to users to explore data cubes at the transaction level.…”