This study investigates the intelligent laser processing of polymethylmethacrylate (PMMA) microchannels, examining the influence of laser power density, scanning speed, and scanning passes on channel dimensions utilizing a CO2 laser. A back propagation (BP) neural network model was developed to predict the laser processing parameters. The model was trained with empirical data and further optimized using a particle swarm optimization algorithm. The results showed that the neural network optimization model can control the PMMA microchannel width machining error within 5% and the depth machining error within 12%. The model has good prediction accuracy and will provide a basis for the intelligent selection of laser processing parameters of PMMA microchannel.