Comparing supervised learning methods for classifying sex, age, context and individual Mudi dogs from barking

摘要: Barking is perhaps the most characteristic form of vocalization in dogs; however, very little known about its role intraspecific communication this species. Besides obvious need for ethological research, both field and laboratory, possible information content barks can also be explored by computerized acoustic analyses. This study compares four different supervised learning methods (naive Bayes, classification trees, $$k$$ -nearest neighbors logistic regression) combined with three strategies selecting variables (all variables, filter wrapper feature subset selections) to classify Mudi dogs sex, age, context individual from their barks. The accuracy models obtained was estimated means $$K$$ -fold cross-validation. Percentages correct classifications were 85.13 % determining 80.25 % predicting age (recodified as young, adult old), 55.50 % classifying contexts (seven situations) 67.63 % recognizing individuals (8 dogs), so results are encouraging. best-performing method following a selection approach. better than they other approaches reported specialized literature. first time that sex domestic have been predicted help sound analysis. shows dog carry ample regarding caller’s indexical features. Our analysis provides indirect proof may serve an important source well.