Intelligent model for assessing the suitability of water based on its chemical and biological parameters

This paper presents a study on the development and testing of a machine learning model for classifying water according to its purpose, based on chemical and microbiological indicators. The aim of the study is to create an accurate and stable model that can effectively determine the categories of water — drinking, industrial, agricultural and technical — based on the analysis of its qualitative characteristics. The gradient boosting method was chosen as the algorithm for classification, which demonstrated high accuracy and reliability when working with multidimensional data. The study included the processing of a synthetic dataset containing information on water parameters such as calcium concentration, total bacterial content, phosphate levels, alkalinity, sulfates and oxidizability. The analysis of the correlation matrix made it possible to identify the relationships between the features, which made it possible to optimize the operation of the model and improve its accuracy. The quality of the model was tested using accuracy, completeness and F1-score metrics, which demonstrated almost perfect results for all categories. In conclusion, it is concluded that the proposed model can be successfully used to automate the analysis of water quality in ecology, industry and agriculture. The use of intelligent technologies opens up new opportunities for monitoring water resources, increasing the efficiency and reliability of quality control.