This research aimed to design a robust mathematical model for forecasting and planning the budget of Tehran Municipality under conditions of economic uncertainty. Its main focus was on identifying indicators affecting revenue and cost and combining multivariate regression with multi-objective optimization. In this study, key indicators were first identified and localized through literature meta-synthesis and fuzzy Delphi method. Then, 9-year real data of revenue, cost and economic indicators of Tehran Municipality were collected and after pre-processing, they were analyzed with different regression models such as Ridge, Lasso, Elastic Net and Random Forest. The best models were selected with ten-way cross-validation and revenue and cost forecasts were presented with a 95% confidence interval. Next, a robust multi-objective optimization model was designed with the Bertsimas-Sim approach and the NSGA-III hybrid algorithm.The objectives of this model included maximizing net profit, reducing deviation from council goals, minimizing regret, and complying with legal constraints. The results showed that the Elastic Net and Lasso models performed very well in forecasting revenue and expenses with high accuracy and a difference of less than 0.3 percent compared to the 1404 supplementary budget. Also, the robust model, by adjusting the appropriate level of protection, provided optimal solutions for budget allocation, in which the transportation sector had the highest priority and the share of other missions was also maintained. Analysis of different scenarios also showed that this model is resistant to economic fluctuations and can be an effective tool for municipal budget decision-making under conditions of uncertainty