Abstract: To address the problem of low accuracy in the blast furnace ironmaking model， a precise multi-objective optimization model for the blast furnace production process was established.Firstly， the massive data from blast furnace were pre-processed before machine learning. Then， six kind of machine learning algorithms were used to predict coke ratio and K value， including support vector machine， random forest， gradient boosting regression tree， XGBoost， LightGBM and artificial neural network.Feature engineering and hyper-parameter tuning were used to optimize the prediction results from machine learning.Finally， the new ensemble learning method was used for prediction.Consequently， the proposed machine learning method has not only high accuracy， but also good robustness.Based on the prediction from machine learning， multi-objective optimization analysis of blast furnace parameters is further carried out by NSGA-Ⅱ algorithm so that Pareto optimal solution can be obtained. Therefore， the blast furnace operator can select the corresponding control parameters， according to these multi-objective optimization results.

Key words: blast furnace (BF), machine learning, ensemble learning, genetic algorithm, parameter prediction