Correlation between Privacy Preserving Data Publishing and Feature Selection Stability

Abstract Data mining is the technique of getting useful information from huge amount of available data stored in organizations. In these days, microdata published are high-dimensional. Feature selection is an important dimensionality reduction technique for data mining. Selection stability is the robustness of feature selection algorithms for the small perturbations of data sets. Selection stability is an important criterion for the feature selection in data mining. Earlier researches have been in the direction that selection stability is algorithmic dependent. But recently researches proved that selection stability is data dependent but not completely algorithmic independent. Privacy preserving data publishing is the publishing of public data to the private parties for research purposes after the perturbation of data to preserve privacy. In privacy preserving data publishing, we modify the data in some way in order to preserve the privacy of the data. This perturbation affects the selection stability and also the utility of the data. This paper finds the relationship between privacy preserving data publishing and feature selection stability.