Microwave-assisted pyrolysis is an advanced route for biomass conversion into bio-oil, biochar, and syngas. Accurate prediction of product yields remains challenging due to complex interactions between feedstock characteristics and operating parameters. In this study, a comprehensive dataset of 227 experimental runs was used to develop a hybrid modeling framework combining Random Forest (RF) regression and a Genetic Algorithm-optimized Adaptive Neuro-Fuzzy Inference System (ANFIS-GA). The catalyst type comprising carbon type, zeolites, metal oxides, and alkali was incorporated as a categorical input variable to account for its influence on microwave absorption and reaction selectivity. The RF model used biomass composition, catalyst type, and operating parameters as inputs but showed limited prediction accuracy (R2  0.98, MAE <1.5, MSE <5, and RMSE <2.3). Sensitivity analysis indicated that ash content predominantly affected char yield, carbon content influenced oil yield, while temperature strongly governed gas formation. These findings are consistent with reported microwave pyrolysis trends. The proposed hybrid framework provides reliable predictive capability and supports data-driven optimization of microwave biomass pyrolysis processes.