In order to improve the level of on-line ash detection in the process of coal slime flotation, this research studied the prediction of tailings ash content in the process of coal slime flotation, and proposed a soft measurement method. The Lightweight Gradient Elevator (LightGBM) algorithm model was used to model the ash prediction, and combined with the genetic algorithm to optimize the parameters, a LightGBM flotation tailings ash prediction model based on genetic algorithm optimization was designed. The model training dataset was constructed by collecting the production process data, including slurry flow, concentration, foaming dose, capture dose and dry slime volume. The experimental results show that the average absolute error of the prediction results of the model is 0.72, which is 11.1% higher than that of the unoptimized LightGBM model, and 15.8% lower than that of the best decision tree model in other models. This further proves the effectiveness of the proposed model in the prediction accuracy of tailings ash content, and provides new technical support for the realization of intelligent flotation.