Revisiting probabilistic neural networks: a comparative study with support vector machines and the microhabitat suitability for the Eastern Iberian chub (Squalius valentinus)

摘要: Abstract Probabilistic Neural Networks (PNNs) and Support Vector Machines (SVMs) are flexible classification techniques suited to render trustworthy species distribution habitat suitability models. Although several alternatives improve PNNs' reliability performance and/or reduce computational costs exist, PNNs currently not well recognised as SVMs because the were compared with standard PNNs. To rule out this idea, microhabitat for Eastern Iberian chub (Squalius valentinus Doadrio & Carmona, 2006) was modelled four types of (homoscedastic, heteroscedastic, cluster enhanced PNNs); all them optimised Differential Evolution. The fitness function criteria (correctly classified instances, true skill statistic, specificity sensitivity) partial dependence plots used assess respectively each model. Heteroscedastic achieved highest in every index but specificity. However, these two rendered ecologically unreliable plots. Conversely, homoscedastic reliable Thus, proved be a eurytopic species, presenting microhabitats cover present, low flow velocity (approx. 0.3 m/s), intermediate depth 0.6 m) fine gravel (64–256 mm). outperformed SVMs; thus, based on results PNN, which also showed high values criteria, we would advocate combination approaches (e.g., heteroscedastic or PNNs) balance trade-off between accuracy