<p>The reinforcement of soft soil subgrades with industrial solid waste has become an economical and environmentally friendly method. However, determining the optimal mix ratio for different combinations of solidifiers is a huge challenge that requires a large number of orthogonal tests to verify. Therefore, this study aims to establish an intelligent mixing optimization method for the ratio scheme of industrial solid waste reinforcement of soft soil subgrade. A dataset considering multiple parameters such as solid waste type, addition ratio, and curing age is established by summarizing literature and laboratory tests. An unconfined compressive strength (UCS) prediction model based on eXtreme Gradient Boosting (XGBoost) is developed. The prediction model demonstrates high evaluation performance, with a correlation coefficient (<i>R</i>²) of 0.965 in training set and 0.891 in testing set. Using the XGBoost prediction model as the fitness function and other parameters as constraint conditions, an optimization mechanism for determining the optimal mixture ratio of specific solidifiers is established based on the global search capability of the particle swarm optimization (PSO) algorithm. The XGBoost-PSO model is tested for single, double, and multiple hardener combinations to verify the applicability of the intelligent ratio optimization method. Based on regulatory requirements, the optimal ratio is verified using alkali slag, steel slag, and mineral slag as hardeners. The research results provide guidance for optimizing the soft soil reinforcement process in roadbed construction and improving resource utilization.</p>