OLAP data cube compression techniques: A ten-year-long history

Abstract

OnLine Analytical Processing (OLAP) is relevant for a plethora of Intelligent Data Analysis and Mining Applications and Systems, as it offers powerful tools for exploring, querying and mining massive amounts of data on the basis of fortunate and well-consolidated multidimensional and a multi-resolution metaphors over data. Applicative settings for which OLAP plays a critical role are manyfold, and span from Business Intelligence to Complex Information Retrieval and Sensor and Stream Data Analysis. Recently, the Database and Data Warehousing research community has experienced an explosion of OLAP-related methodologies and techniques aimed at improving the capabilities and the opportunities of complex mining processes over heterogeneous-in-nature, inter-related and massive data repositories. Despite this, open problems still arise, among which the so-called curse of dimensionality problem plays a major role. This problem refers to well-understood limitations of state-of-the-art OLAP data processing techniques in elaborating, querying and mining multidimensional data when data cubes grow in size and dimension number. This evidence has originated a large spectrum of research efforts in the context of Approximate OLAP Query Answering techniques, whose main idea consists in compressing target data cubes in order to originate compressed data structures able of retrieving approximate answers to OLAP queries at a tolerable query error. This research proposes an excerpt of a ten-year-long history of OLAP data cube compression techniques, by particularly focusing on three major results, namely Δ - Syn, K LSA and ℒCS - Hist. © 2010 Springer-Verlag Berlin Heidelberg.