Predicting shrimp disease occurrence: artificial neural networks vs. logistic regression

Predicting the occurrence of disease outbreaks in aquacultural farms can be of considerable value to the long-term sustainable development of the indu stry. Prior research on disease prediction has essentially depended upon tr aditional statistical models with varying degrees of prediction accuracy. F urthermore, the application of these models in sustainable aquaculture deve lopment and in controlling environmental deterioration has been very limite d. In an attempt to look for a more reliable model, we developed a probabil istic neural network (PNN) to predict shrimp disease outbreaks in Vietnam u sing farm-level data from 480 Vietnamese shrimp farms, including 86 semi-in tensive and 394 extensive farms. We also compared predictive performance of the PNN against the more traditional logistic regression approach on the s ame data set. Disease occurrence (a 0-1 variable) is hypothesized to be aff ected by a set of nearly 70 variables including site characteristics, farmi ng systems, and farm practices. Results show that the PNN model has a bette r predictive power than the logistic regression model. However, the PNN mod el uses significantly more input (explanatory) variables than the logistic regression. The logistic regression is estimated using a stepwise procedure starting with the same input variables as in PNN model. Adapting the same input variables found in the logistic regression model to the PNN model yie lds results no better than the logistic regression model. More importantly, the key factors for prediction in the PNN model are difficult to interpret , suggesting besides prediction accuracy, model interpretation is an import ant issue for further investigation. (C) 2000 Elsevier Science B.V. All rig hts reserved.