Warianty tytułu
Języki publikacji
Abstrakty
This document describes the winning method for the AAIA'14 Data Mining Competition: Key risk factors for Polish State Fire Service. The competition challenge was a feature selection problem for a set of three classifiers, each of them in a form of ensemble of naive Bayes classifiers. The method described in this paper uses a genetic algorithm approach to identify an optimal set of variables used by the classifiers. The optimal set of variables is found through a three-stage procedure that involves different settings for the genetic algorithm. The first step leads to reduction of attribute set under consideration from 11,582 to 200 attributes. The following two steps focus on finding an optimal solution by first exploring the solution space and then refining the best solution found in an earlier step.(original abstract)
Rocznik
Tom
Strony
381--385
Opis fizyczny
Twórcy
autor
- Cranfield University
Bibliografia
- Dietterich T., Overfitting and undercomputing in machine learning, ACM Comput. Surv. 27, (3), 326-327, 1995
- Geisser S., "The predictive sample reuse method with applications", J. Amer. Statist. Assoc., 70:320-328, 1975
- Holland J. H., "Adaptation in natural and artificial systems", Ann Arbor: The University of Michigan Press, 1975
- Mitchell M., "An Introduction to Genetic Algorithms", MIT Press, 1998
Typ dokumentu
Bibliografia
Identyfikatory
Identyfikator YADDA
bwmeta1.element.ekon-element-000171325199