To address the issue of time consumption and labor intensity in traditional testing methods for the basic performance of rubber asphalt, a performance prediction method based on BP neural networks was proposed. A large amount of experimental data was normalized and analyzed, and standardization techniques were utilized to enhance the model's training efficiency. An artificial neural network (ANN) prediction model was constructed for the penetration, softening point, and ductility of rubber asphalt. Genetic algorithms were applied for optimization during the model training process, and early stopping and learning rate scheduling were employed to ensure the model's generalization ability. The results indicate that the evaluation of the test set demonstrates good performance in prediction, with a root mean square error (RMSE) of 0.63, a mean absolute error (MAE) of 0.45, and a correlation coefficient (R) of 0.91. This model effectively predicts the basic properties of rubber asphalt, improves prediction accuracy, and provides a reliable reference for future predictions of rubber asphalt properties.