Ant colony optimization: A powerful strategy for biomarker feature selection

Weixiang Zhao, Cristina E Davis

Research output: Chapter in Book/Report/Conference proceedingChapter

Abstract

As instrumentation develops in industry and science, we frequently generate multi-dimension data sets that may involve input from a large number of factors or variables. Many parameters of these instrument systems may not be directly related with the core function of the systems, and some factors may even lead to noise contamination of output signals. The potential obscuring effects of these variables on the data set can make it difficult to determine which parts of the instrument data are the most meaningful. Therefore, feature selection within data sets is becoming a core technique to detect pertinent factors or variables for system characterization. This not only reduces the data dimension but also provides pertinent information for system mechanism studies, and can ultimately yield information about the underlying instrumentation function. Feature selection within data sets has been attempted using a variety of different methods, and some conventionally used methods include statistical analyses such as Student's t-test, the Fisher-ratio and analysis of variance (ANOVA); however, these methods may not always be feasible for nonlinear systems and non-classification problems. As an artificial intelligence method, genetic algorithm may provide a novel feature selection strategy to detect pertinent features for a variety of systems, even for those without clear mechanisms. And this type of biological inspired adaptive learning method has prompted other new approaches in feature selection, such as the ant colony algorithm (ACA) method. The ant colony algorithm that mimics the social behavior of ants is a typical swarm intelligence based optimization method, and this approach has increasingly been applied for system feature selection. This commentary will provide a short review of recent ACA based feature selection studies, compare the outcomes of these studies to other intelligent feature selection methods, and discuss the advantages and disadvantages of the ACA based feature selection method. Together this chapter will suggest promising directions for future research in this area.

Original languageEnglish (US)
Title of host publicationApplications of Swarm Intelligence
PublisherNova Science Publishers, Inc.
Pages193-197
Number of pages5
ISBN (Print)9781617286025
Publication statusPublished - 2011

    Fingerprint

ASJC Scopus subject areas

  • Engineering(all)

Cite this

Zhao, W., & Davis, C. E. (2011). Ant colony optimization: A powerful strategy for biomarker feature selection. In Applications of Swarm Intelligence (pp. 193-197). Nova Science Publishers, Inc..