A review of Feature Selection Stability
Abstract
Business establishments for getting tactical information on their day to day business activities use data mining as an indispensable technique. For data mining applications, feature selection is crucial as these days’ microdata is mostly high-dimensional. The sturdiness of feature selection algorithms in a succeeding rehearsal of feature selection experiments even for small perturbation of microdata is called feature selection stability. Selection stability considered one of the indispensable criteria for feature selection algorithms by the researchers. This paper describes selection stability, the data perspective nature of selection stability along with selection stability measures. The selection stability is related to privacy conserving data mining as the selection stability is generally data reliant which is also explained in this paper.