Feature Subset Selection is a well known task in the Machine Learning, Data Mining, Pattern Recognition and Text Learning paradigms. In this chapter, we present a set of Estimation of Distribution Algorithms (EDAs) inspired techniques to tackle the Feature Subset Selection problem in Machine Learning and Data Mining tasks. Bayesian networks are used to factorize the probability distribution of best solutions in small and medium dimensionality datasets, and simpler probabilistic models are used in larger dimensionality domains. In a comparison with different sequential and genetic-inspired algorithms in natural and artificial datasets, EDA-based approaches have obtained encouraging accuracy results and need a smaller number of evaluations than genetic approaches.
For the entire collection see [Zbl 0979.00024].