ABSTRACT
This work presents the creation of a neural network trained using a dataset from SEMASA (Serviço Municipal de Água, Saneamento Básico e Infraestrutura) from Itajaí, containing pH, turbidity, and apparent color data of water bodies from various neighborhoods in the municipality, capable of classifying water samples according to the water types defined by the CONAMA (Conselho Nacional do Meio Ambiente) 357/05 standard for embedding in an IoT device. The dataset is composed of 8,928 records from 2018 to 2023, covering neighborhoods São Roque I and II, Arapongas, and Limoeiro. The selected variables are critical for water quality assessment, where pH indicates the acidity or alkalinity of the water, turbidity measures the presence of suspended particles, and apparent color is a visual indicator of dissolved and suspended materials.