Abstract <p>Freshwater is a vital natural resource, yet its contamination by toxic waste from urban areas and industries often renders it unsuitable for consumption by living beings. This research advances a novel ANOVA-based hypothesis-driven machine learning (ML) framework designed for the monitoring and regulation of water quality in rivers situated within the state of Uttar Pradesh (UP), India. Concentrated efforts entail the collection of water specimens from designated sampling sites within the Water Quality Monitoring System (WQMS). Through the application of linear regression analysis, an appraisal of the relationships and congruence across multiple physicochemical parameters is undertaken. The dataset undergoes stratification into training and testing subsets, distributed at proportions of 25%, 50%, and 75%. Preceding this partitioning, the water weight values embedded within the data samples are extracted. After this preparatory phase, an ANOVA examination is conducted, delivering insights into the coefficient of correlation, <i>p</i>-value, <i>F</i>-value, and standard error inherent to the resultant model. Significantly, the model yields an impressive 90% accuracy during its assessment against the dataset. This achievement underscores its efficacy in the domain of water quality evaluation and management.</p>