Feature Selection: Multi-source and Multi-view Data Limitations, Capabilities and Potentials

Abstract

Feature Selection (FS) is a crucial step in high-dimensional and big data analytics. It mitigates the `curse of dimensionality' by removing redundant and irrelevant features. Most FS algorithms use a single source of data and struggle with heterogeneous data, yet multi-source (MS) and multi-view (MV) data are rich and valuable knowledge sources. This paper reviews numerous, emerging FS techniques for both these data types. The major contribution of this paper is to underscore uses and limitations of these heterogeneous methods concurrently, by summarising their capabilities and potentials to inform key areas of future research, especially in numerous applications.