Abstract: Control of blast furnace ironmaking process means to control the temperature and compositions of blast furnace hot metal within specific bounds. For the silicon content in blast furnace hot metal as a targeted indicator of in-furnace thermal state, a data-driven based predictive model is constructed in this paper utilizing the information of multivariate time series measured from the blast finance system. Through an example, it is indicated that the constructed data-driven based predictive model has a good performance in predicting the silicon content in hot metal, with 83.23% percentage of target hitting and 0.07260 of the root mean square error of prediction when the size of predicted sample set is 167. These criteria are better than those achieved by data-driven predictive model based on univariate time series, implying that the constructed data-driven predictive model based on multivariate time series can act as an important tool to serve the production process.