<p>The requirement for engineering components with superior surface quality and precise dimensional accuracy has increased significantly in recent years. To fulfil these requirements, a belt-type magnetorheological finishing (BT-MRF) technique, based on the principles of magnetorheological finishing&#xa0;(MRF), has been introduced. The technique is intended to finish large components’ external surfaces made of both ferrous and non-ferrous materials. The suggested technique produces ultra-precision surfaces on materials utilized in a variety of industrial applications and enhances finishing performance. An experimental investigation has been conducted to optimize machining parameters and investigate their impacts in order to achieve a maximum percentage change in average surface roughness (%Δ<i>R</i>_<i>a</i>). Response surface methodology (RSM) and artificial neural network (ANN), a machine learning technique, have been employed to evaluate the experimental results. The predictive performance of a second-order quadratic model developed by RSM is excellent, with <i>R</i>² adjusted <i>R</i>² and predicted <i>R</i>² values of 0.9662, 0.9347, and 0.8433, respectively. With an overall regression value of 0.97302, ANN demonstrates outstanding prediction accuracy. Comparative evaluation confirms that ANN offers superior predictive capability, while the combined application of response surface analysis and ANN provides an effective optimization framework for the BT-MRF process.</p>