Modeling and predicting caffeine contamination in surface waters using artificial intelligence and standard statistical methods

Caffeine, considered an emerging contaminant, serves as an indicator of anthropic influence on water resources. This research employs various modeling techniques, including Artificial Neural Networks (ANN), Random Forest (RF), and more, along with hybrid and ensemble methods, to predict caffeine concentrations (in regression and classification scenarios) using readily available water quality parameters. The results indicate Ensemble-RF as the most effective method for estimating caffeine concentrations, while classification scenarios highlight Ensemble-RF, ANN, and Ensemble-ANN as promising methodologies for predicting contamination levels. This study offers a valuable tool for swiftly assessing caffeine contamination in water, leveraging easily obtainable data, with implications for safeguarding water resource systems.