Abstract: To address the problem of insufficient accuracy in the application of ground-penetrating radar for coal-rock interface prediction, the Particle Swarm Optimization (PSO) algorithm was used to optimize the BP neural network, and a coal-rock interface prediction model based on ground-penetrating radar and PSO-BP neural network was established. The single-sided reflection method of ground-penetrating radar was employed to detect the coal-rock interface, and the radar image response characteristics under different conditions were summarized to determine the coal-rock interface characteristic parameters, including coal proportion, amplitude at the response position, average amplitude at coal response position, amplitude attenuation value, two-way travel time of reflection wave, electromagnetic wave velocity, and coal dielectric constant. Based on the selected characteristic parameters, dielectric constant tests and simulated coal-rock interface recognition experiments were carried out to obtain measured sample data. The PSO algorithm was used to optimize the weights and thresholds of the BP neural network to obtain the optimal model. The coal-rock interface characteristic parameters were then input into the PSO-BP neural network model to predict the coal-rock interface. The experimental results showed that, compared with GA-BP and BP neural network models, the MSE of the PSO-BP model decreased by 22.14% and 45.54%, the MAPE decreased by 22.22% and 46.15%, and the MAE decreased by 31.58% and 55.68%, respectively. The PSO-BP model has significant advantages in prediction accuracy, error control ability, and data fitting performance, predicting coal-rock interface positions closer to the actual locations with better stability.