<p>With greenhouse gas (GHG) emissions from solid waste disposal sites being a major environmental concern, their accurate prediction is required for the betterment of policy analysis and interventions. However, the task is challenging due to complex nonlinear interactions and high spatiotemporal variability in the emission processes. This article addresses the pertinent research gap through the evaluation of the potential of three alternate machine learning regression approaches, namely classical multiple linear regression (MLR), genetic algorithm (GA)-based evolutionary modelling approach and artificial neural networks (ANN, specifically a multilayer perceptron or MLP). Accordingly, the prediction and forecasting of the GHG emissions from solid waste incineration and landfilling systems in Guwahati, India, has been addressed using data from 1970 to 2018. Methane (CH₄), carbon dioxide (CO₂) and particulate matter (PM_2.5 and PM₁₀) fluxes were analyzed with the statistical comparison and cross-validation techniques. Among the tested models, the MLP outperformed others and achieved R² values of 0.99 (training) and 0.97 (prediction), RMSE of 0.32 and an Index of Agreement (IoA) of 0.94. All these convey a strong alignment between predicted and observed values. While the GA model also demonstrated competitive results (R² = 0.95; RMSE = 0.63; IoA = 0.89) and showcased robustness and adaptability, the MLR model captured only linear trends. The superior predictive accuracy of the MLP underscores its potential for the reliable prediction of CH₄ (24.8&#xa0;g/m²/day) and CO₂ (0.40&#xa0;g/m²/day) fluxes and PM concentrations (R² = 0.99 for PM_2.5; 0.98 for PM₁₀). These findings, through the demonstration of neural network-based approaches in the field of environmental modelling, enhance the competence to forecast the GHG emissions and can consequently improve strategies adopted for data-driven waste management and climate mitigation.</p>