Abstract
Most feature selection methods identify only a single solution. This is acceptable for predictive purposes, but is not sufficient for knowledge discovery if multiple solutions exist. We propose a strategy to extend a class of greedy methods to efficiently identify multiple solutions, and show under which conditions it identifies all solutions. We also introduce a taxonomy of features that takes the existence of multiple solutions into account. Furthermore, we explore different definitions of statistical equivalence of solutions, as well as methods for testing equivalence. A novel algorithm for compactly representing and visualizing multiple solutions is also introduced. In experiments we show that (a) the proposed algorithm is significantly more computationally efficient than the TIE* algorithm, the only alternative approach with similar theoretical guarantees, while identifying similar solutions to it, and (b) that the identified solutions have similar predictive performance.
Highlights
Feature selection is an essential part of data analysis tasks which focus on knowledge discovery and improving understanding of the problem under study
While finding a single solution may be acceptable for building a predictive model, it is not sufficient when feature selection is employed for knowledge discovery
In order to reduce the chance of false positive equivalences, we recommend to (a) perform extensive tuning of the hyper-parameters3 of the feature selection algorithm, in order to increase the chance of identifying Markov blankets, (b) first apply a permutation-based variance test for PEQ or MEQ to quickly filter out false equivalences and (c) afterwards apply an IEQ test using the comprehensive approach to decide for equivalence, and (d) use relatively high significance levels to further reduce the number of false positives
Summary
Feature selection is an essential part of data analysis tasks which focus on knowledge discovery and improving understanding of the problem under study This is no accident, as the solution has been shown to be directly related to the data-generating causal mechanism (Koller and Sahami 1996; Tsamardinos and Aliferis 2003; Aliferis et al 2010). The difference in predictive performance between two solutions may not be statistically distinguishable In domains such as molecular biology there often exist multiple solutions, possibly because of the inherent redundancy present in the underlying biological system (Dougherty and Brun 2006; Statnikov and Aliferis 2010). A feature selection algorithm should identify all solutions that are “equivalent” (for some reasonable definition of equivalence) Another advantage of outputting multiple solutions is that one could use any of them for building a predictive model.
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Disclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.