Lignocellulosic agro-waste is a renewable, potential, and sustainable feedstock for the production of plastic precursors like 2,5-furandicarboxylic acid (FDCA). One-pot conversion of sugarcane bagasse into FDCA is a promising, cost-effective method to valorize residual biomass and improve process sustainability by reducing environmental pollution. In this work, a heterogeneous catalyst from environmentally benign and non-precious metal oxides of iron and manganese on zeolite molecular sieves 5A support (FMZ catalyst) was developed for one-pot conversion of bagasse to FDCA. Optimization and prediction models for FDCA yield and selectivity for the process were developed by employing Box-Behnken Design (BBD) of Response surface methodology (RSM) and machine learning model, i.e., Artificial Neural Network - Levenberg Marquardt (ANN-LM). Process variables viz. time, temperature, and catalyst dosage were selected as the factors, while FDCA yield and selectivity were the responses. The same data set was used for predictive modeling with ANN-LM model. The RSM approach achieved a model R