Use of data-driven model to analyse the occurrence patterns of an indicator fish species in river : a case study for Alburnoides eichwaldii (De Filippi, 1863) in Shafaroud River, north of Iran

present study aims to integrate multinomial logistic regression with an input variable selection method, genetic algorithm, GA, to select the most important explanatory variables for evaluating the occurrence patterns of the bleak (Alburnoides eichwaldii) in river. Seven different sampling sites (from the source to the mouth of the Shafaroud River, north of Iran) were considered to analyse the probability of occurrence of the fish during one year sampling campaign. The abundance of bleak (based on 42 fish presence and 42 fish absence data, as outputs of model) together with a set of physical-chemical water characteristics and river morphology (84 instances as inputs of model) were monthly and repeatedly recorded at each sampling site. Two-third of instances (56) was used for training and the remaining of instances (28) as test set. The results of paired Student's t-test showed that the predictive performances of model (% correctly classified instance and Kappa statistics) were improved after variable selection method. GA selected 9 of 18 input variables including dissolved oxygen, pH, water temperature, river depth, electric conductivity, total hardness, nitrite, orthophosphate and sulphate. The curves of binary logistic regression confirmed that increasing three of the selected variables (dissolved oxygen, water temperature and pH) might increase the probability of bleak presence while increasing concentration of other selected variables might decrease the probability of fish occurrence in the river basin (p < 0.05 for all). Information on the occurrence patterns of indicator fish species in relation to physico-chemical water quality and habitat variables is important to develop conservation and management programs for the study river, where water resources are facing with high risk of pollution due to increasing of agricultural and industrial development activities