Ecological Systems and Devices Annotation << Back CLASSIFICATION OF WATER SOURCE STATE USING MACHINE LEARNING V.N. Klyachkin, I.N. Karpunina A method for classifying the states of a water source based on the results of monitoring the physicochemical parameters of water using the "random forest" method in the Statistica package has been developed. The features of the selection of parameters of this method are shown, the results of forecasting by the proportion of errors on the training and test samples, as well as by the classification matrix, are analyzed. The study was conducted on the example of the water treatment system at the station for the purification of natural surface water and the preparation of drinking water in St. Petersburg (water source "Western Kronstadt"). Keywords: water treatment system, water source status, multiclass classifi cation, machine learning, random forest, classification matrix. DOI: 10.25791/esip.7.2022.1307 Pp. 20-25.