A comparative analysis of feature selection stability measures

Now-a-days, data mining become indispensable for business organizations for decision making, which makes use of the information from huge amount of archived data. Due to the advancements in information technology, there will be proliferation of extremely high-dimensional data. Feature selection manages the "curse of dimensionality" as it is an important dimensionality reduction technique. Recently, Stability or robustness of feature selection methods becomes a hot topic of interest for researchers. Feature selection stability is the measure of the sensitivity of feature selection algorithms for the slight perturbations in the experimental dataset. There are various selection stability measures which have been used to measure the stability of feature selection algorithms based on the result sets. This paper gives an account of various selection stability measures and also the merits and demerits of each stability measure that have been explored using a set of experimental datasets.