In the past years, there has been a lot of interest in water quality and its prediction as there are many pollutants that affect water quality. Surface (S) water quality (WQ) monitoring is among the best environmental goals in light of the current droughts and the growth of both urban and rural areas. The prediction and management of water quality are critical to ensure sustainable water resources, particularly in regions like Mahanadi, Odisha, where rivers face increasing pollution from industrailization, agriculture, and urban expansion. The techniques provided herein will help us in controlling and reducing the risks of water pollution. A particular focus is given to hybrid models that integrate multiple approaches to improve predictive accuracy and robustness. Therefore, the purpose of the work was to map SWQ in the Mahanadi River Basin, Odisha, by utilizing Fuzzy Logic (FL) drinking WQ index (I), and validating machine learning (ML) algorithms using Group Method of Data Handling (GMDH) and Ridge (R) theory. On this basis, in the current study, physicochemical parameters related to WQ, including 20 indicators, were interpolated at 19 sampling sites, for the monsoon period, of duration 2018–2024. The results were compared against the World Health Organization (WHO) drinking water standards to assess compliance and identify areas affected by domestic, agricultural, and industrial discharges. The physicochemical distribution results established that greater amounts of Cl−, SO42−, Coliform and Nitrogen, detected on coastal regions, indicating the existence of salt buildup and seawater intrusion. Findings from the FL-DWQI assessment revealed that surface water quality throughout the research region, varied from excellent to unsuitable, indicating that most water quality locations did not meet the standard limits set by the WHO. Overall, the score ranged between 19 and 441, in the study area. Thirteen sampling stations displayed highest score, whereas sampling station SN-(3) showed the lowest score. Based on the G-WQI map creation, 47.37% of the investigated area had fallen into poor condition. The percentage of the classes in this area, with average and good WQ was 21.05 and 31.58%, respectively. Similarly, the R-WQI based on the Ridge technique, indicated that most of the sites, specifically 68.42%, were classified as poor water quality. Consequently, only 5.26 and 26.32% of the locations were identified as medium and good water quality. However, the R-WQI values reported the unsuitability water with a percentage of around 94.74%. Results show that around 68.42% (FL-DWQI), 47.37% (G-WQI) and 42.11% (R-WQI) of samples observed to be of deteriorated water quality during the study period. Findings highlight the impact of both point and non-point source pollution, with several sites exceeding acceptable limits for key contaminants. The findings indicate that all three models provide accessible and understandable maps portraying the water quality in the research area. As a result, the conclusions emphasize the importance of advanced machine learning models in river water modelling applications, giving useful insights for water resource management and decision making. Therefore, the study underscores the need for continuous water quality monitoring and the implementation of targeted municipal regulations to mitigate environmental degradation and promote long-term ecological sustainability. This work identifies key research gaps and proposes future directions for developing transparent, adaptive, and accurate models. The findings can also guide researchers and policymakers towards the development of smart water quality management systems that enhance decision-making and ecological sustainability.