ARTIFICIAL INTELLIGENCE IN ECOTOXICOLOGY: PREDICTIVE MODELS FOR CHEMICAL IMPACT ON WILDLIFE

Abstract

 The paper examines the use of artificial intelligence in ecotoxicology with an emphasis on the use the models to mitigate the effects of polluted chemicals on wildlife. In particular, upon having several data sets with information on different chemical substances and their impact on different species, we applied four machines learning algorithms: RF, SVM, NN, and GBM. This work showed that the RF model had the highest level of accuracy of 92% while GBM had 89%, NN had 87%, and SVM had 85% of correctly predicting toxicological outcomes. These machine learning models are substantially more accurate than conventional approaches, thus their capability for giving accurate, quantitative descriptions of chemical toxicity. Observing the comparative analysis with the prior research showed that the AI models provided lower prediction errors on the range of 5-15% and provided enhanced interpretability of the multifaceted data sets. The implications of this research are that AI should be adopted for use in environmental risk assessment particularly for wildlife protection. Future studies should also aim at making finer nuances of these models and adding versatility to these models with respect to ecological settings.