Dimension Reduction and Feature Selection

Abstract

Data Mining algorithms search for meaningful patterns in raw data sets. The Data Mining process requires high computational cost when dealing with large data sets. Reducing dimensionality (the number of attributed or the number of records) can effectively cut this cost. This chapter focuses a pre-processing step which removes dimension from a given data set before it is fed to a data mining algorithm. Thiswork explains howit is often possible to reduce dimensionality with minimum loss of information. Clear dimension reduction taxonomy is described and techniques for dimension reduction are presented theoretically.