In order to provide qualified sinter for the blast furnace, a large data method combining XGBoost algorithm, factor correlation analysis and deep learning algorithm was proposed to predict the sinter content of less than 10 mm based on a large amount of data in each part of the sintering production. Firstly, the data from the sinter plant database need to be collected, integrated and preprocessed. Secondly, the factor analysis was carried out, to select 14 relevant variables suitable for model training and to perform correlation analysis between variables. Finally, the deep neural network algorithm model was established. By testing the model and comparing with the traditional algorithm model, the results showed that the model prediction effect was very good, and the purpose of accurately predicting the content of sinter particle size less than 10 mm was achieved, which had a good guiding significance for the actual production of sintering.