Under-Updated Particle Swarm Optimization for Small Feature Selection Subsets from Large-Scale Datasets

Feature selection is the process of choosing a subset of features from the original larger set that are related to a certain outcome, such as disease type, dose, income, and time to event. The use of feature selection procedures is almost compulsory and complex in biology and medicine because the generation of massive datasets is nowadays common for many stateof-the-art technologies such as transcriptomics, proteomics, metabolomics, and genomics where a single, conventional, and relatively cheap experiment may yield the measurement of several thousands of features per sample (Hieter and Boguski, 1997; Sauer et al., 2005). In such cases, feature selection is used to reduce complexity and large computational costs, as well as to improve pattern recognition accuracy, data interpretability and hypothesis generation (Shen et al., 2008; Vapnik, 1998; Guyon et al., 2002).