Controlling subgrade settlement is essential in highway engineering. Subgrade settlement has a linear and nonlinear relationship with its influencing factors. Over-fitting easily occurs in traditional neural network modeling in the presence of numerous input independent variables and results in the low prediction accuracy of the network model. This work aims to address these issues. Thus, the ability of the genetic algorithm to optimize the weight and threshold of the neural network is investigated, and the influence of the set of genetic parameters on the output results is discussed. Experiments with the proposed method show that the optimized BP neural network has higher prediction accuracy and better prediction effect than the traditional neural network model in the simulation of measured data for the Chengdu-Nanchong Highway. The prediction method can be used as an effective auxiliary means for predicting the long-term settlement of highway subgrades.