The precise prediction of nitrogen contentNitrogen content in the steelmaking process of high-nitrogen stainless steelStainless steel has a significant impact on product quality. In the paper, the locally linear embeddingLocally Linear Embedding (LLE)—random forestRandom forest (RF) model has been proposed to predict the nitrogen contentNitrogen content for an 80-ton converterConverter. The thermodynamicThermodynamics and kinetic mechanisms of nitrogen dissolution in the converterConverter are used as a guide. The LLE algorithm is applied for dimensionality reduction and feature extraction. Five machine learningMachine learning models, including extreme learning machine (ELM), random forestRandom forest (RF), gradient boosting decision tree (GBDT), support vector machine (SVM), and back propagation neural networks (BPNN), are utilized to establish nitrogen contentNitrogen content prediction modelsPrediction model. The RF model, which achieved the highest hit ratio, is selected for further modeling. After optimizing the model's hyperparameters, the model is tested using actual production data. The prediction accuracy of the model achieves a hit ration of 91.9% within a deviation range of ±0.015%.