Machine learning-based (ML) models have gained significant prominence in the environmental areas, especially in the wastewater treatment processes. Wastewater treatment plants (WWTPs) are complicated facilities that involve various biological, chemical, and physical processes and are highly subject to uncertainty. Traditional models struggle to capture the nonlinearities in these systems, but ML-based models can model these interactions effectively, which results in the rise in the use of ML models for WWTP modeling and optimization and to provide fast results. ML-based models offer numerous advantages over traditional physical models in predicting output parameters in sewage treatment plants. They can adapt and learn from data, making them particularly valuable when system dynamics are subject to change. Although there are numerous benefits, the utilization of machine learning for modeling sewage treatment plants remains largely unexplored within the Indian subcontinent, especially the comparison of two node-based models, i.e., Random Forest (RF) and artificial neural networks (ANN) in terms of output prediction of the WWTPs. Thus, this study presents ML-based approach for forecasting WWTP’s output variables as a black box by explaining the relationships between an existing WWTP’s different influent and effluent pollution variables for Indian tropical conditions. The study focuses on applying RF and ANN models to predict the outflow variables of wastewater treatment plants. ANN and RF models are developed to deal with number of issues of traditional mathematical models. The developed models concentrate on offering a flexible, useful, and a different approach to model the WWTP’s performance. They are based on data collected from two WWTPs, which include biochemical and chemical oxygen demands, phosphates (PO4−3), and nitrates (NO3−) over a year and three years, respectively. The results indicate high predictive accuracy and generalization capabilities of ML-based models, with ANN achieving the highest correlation coefficient (R2) of up to 0.9025 and RF up to 0.947 for output BOD and demonstrating the efficacy of these machine learning techniques in modeling and managing WWTP processes.