Modeling zero-inflated explanatory variables in hybrid Bayesian network classifiers for species occurrence prediction. (August 2016)

• Record Type: Journal Article
• Authors: Maldonado, A.D.; Aguilera, P.A.; Salmerón, A.
• Is Part Of: Environmental modelling & software. Volume 82(2016:Aug.)
• Page Start: 31
• Page End: 43
• Publication Date: 2016-08
• Subjects: Hybrid Bayesian networks -- Mixtures of truncated exponentials -- Zero excess treatment -- Species distribution modeling; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2016.04.003
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Datasets with an excessive number of zeros are fairly common in several disciplines. The aim of this paper is to improve the predictive power of hybrid Bayesian network classifiers when some of the explanatory variables show a high concentration of values at zero. We develop a new hybrid Bayesian network classifier called zero-inflated tree augmented naive Bayes (Zi-TAN) and compare it with the already known tree augmented naive bayes (TAN) model. The comparison is carried out through a case study involving the prediction of the probability of presence of two species, the fire salamander ( Salamandra salamandra ) and the Spanish Imperial Eagle ( Aquila adalberti ), in Andalusia, Spain. The experimental results suggest that modeling the explanatory variables containing many zeros following our proposal boosts the performance of the classifier, as far as species distribution modeling is concerned. Highlights: A new Bayesian network (BN) classifier for explicitly modeling zero-inflated variables is proposed. Our proposed methodology improves the performance of the classifier regarding a case study involving species distribution. To our knowledge this is the first article that uses BNs to model explanatory variables with zero excess.