Accurate prediction of molten steelSteel composition at the LF refiningLF refining endpoint is the key to improving the quality of high-end steelSteel products. In the paper, a mechanistic and data hybrid modelMechanistic and data hybrid model was proposed for predicting Si and Mn contentsMn content at the LF refiningLF refining endpoint. First, the mechanistic models were constructed based on metallurgical principles to clarify the key influence parameters. Then, the genetic algorithm (GA) was introduced to optimize the hyperparameters of the backpropagation neural network (BPNN), and a data-driven model was used to solve the key parameters of the mechanistic model, achieving the integration of mechanism constraints and data optimization in modeling. The production big data in a plant was used as samples to validate the model. The results showed that the hit rates of Si and Mn contentsMn content in the range of ±0.005% were 93.2 and 91.2%, respectively, which significantly improved the prediction accuracy.