CLASSIFICATION OF DIETARY HABITS OF CHIROPTERANS

Abstract

Bats are the second most successful and abundant mammals on earth belonging to a special order called chiroptera. Chiropterans were captured via mist nets in different agro-ecosystems for identification and morphometric measurements. Ectoparasites were subsequently removed from captured bats and identified. The GLM technique was then employed to investigate and describe the variation in ectoparasite abundance relative to host characteristics under different distributions for count data. The ZIP distribution proved to be the most appropriate distribution for modeling specific and total ectoparasite abundance based on information- theoretic approaches. The findings generally revealed that host Sex, Diet, and Patagial area had strong effects on ectoparasite loads of bats. Furthermore, all identified bats were placed into two foraging guilds (insectivores and frugivores) in order to explore the classification potential of two statistical techniques. Thus, models based on an ANN and BLR were compared in their ability to differentiate between the two dietary habits. The area under the ROC curves were respectively 99.00% and 99.09% for BLR and ANN models. Also, specificity values were 95.30% for BLR and 93.80% for ANN models. The BLR model had a higher sensitivity of 97.60% compared to 96.80% for the ANN model. Although both models had similar diagnostic performance, the BLR model was superior to the ANN model. Hence the addition of other dietary habits as well as predictors should be considered, to help improve the predictive accuracies of the models since both yield acceptable results during classification.