Forecasting Toxic Metal Concentrations in an Inland Sea Ecosystem with Machine Learning Algorithms

Accurately measuring element concentrations within ecosystems is vital for effective environmental monitoring, but these processes are often time-consuming and expensive. When continuous monitoring is required, the associated costs can escalate significantly. In contrast, in recent years, the use of statistical and alternative modeling methods for predicting element concentrations has been increasing. Furthermore, by integrating feature selection techniques into these modeling methods, these approaches enable the identification of key features strongly associated with the target feature, thereby improving model generalization and reducing complexity. This not only improves the prediction performance but also provides significant savings in both time and cost. This study uses machine learning models to explore relationships between Aluminum (Al) and other elements and estimate Al concentration levels in an inland sea ecosystem. Our findings demonstrate that in predicting element concentrations in sea ecosystems, machine learning methods are effective when combined with feature selection.