In this study, a machine learning approach using the Isolation Forest, one of the anomaly detection machine learning algorithms, was proposed to exclude anomalous values from the training data. In addition, the influences of the modified training data on the prediction of airborne chloride were investigated. Therefore, it was found that combining statistical processing with the Isolation Forest improves the accuracy of predicting airborne chloride. Furthermore, it was revealed that the most contributing feature importance to the prediction of airborne chloride is the significant wave height.
