About this book

Feature selection (FS) refers to the problem of selecting those attributes that are most predictive of a given problem, which is encountered in many areas such as machine learning, pattern recognition, systems control and signal processing. FS intends to preserve the meaning of selected attributes; this forms a sharp contrast with those approaches that reduce problem complexity by transforming the representational forms of the attributes.

Feature selection techniques have been applied to small and medium-sized datasets in order to locate the most informative features for later use. Many FS methods have been developed, and this book provides a critical review of these, with particular emphasis on their current limitations. The book systematically presents the leading methods reviewed in a consistent algorithmic framework. The book also details those computational intelligence-based methods (e.g. fuzzy rule induction and swarm optimization) that either benefit from joint use with feature selection or help improve the selection mechanism itself.

From this background, the book first introduces the original approach to feature selection using conventional rough set theory, exploiting the rough set ideology in that only the supplied data and no other information is used. Based on demonstrated applications, the book reviews the main limitation of this approach in the sense that all data must be discrete. It then proposes and develops a fundamental approach based on fuzzy-rough sets, and also presents optimizations and extensions of this approach whose underlying ideas are generally applicable to other FS mechanisms.

Real world applications, with worked examples, are provided that illustrate the power and efficacy of the feature selection approaches covered. In particular, the algorithms discussed have proven to be successful in handling tasks that involve datasets containing huge numbers of features (in the order of tens of thousands), which would be impossible to process further. Such applications include web content classification, complex systems monitoring and algae population estimation. The book shows the success of these applications by evaluating the algorithms statistically with respect to the existing leading approaches to the reduction of problem complexity.

Finally, the book concludes with initial supplementary investigations to the associated areas of feature selection, including rule induction and clustering methods using hybridizations of fuzzy and rough set theories. This opens up many new frontiers for continued research and development of the core technologies introduced.

R. Jensen and Q. Shen. Computational Intelligence and Feature Selection: Rough and Fuzzy Approaches. IEEE Press/Wiley & Sons.