Since the hot metal silicon content simultaneously reflects the product quality and the thermal state of the blast furnace, its modeling is crucial and representative. Although many models have been proposed, there is little research on the inputs of hot metal silicon content model. Accurate selection of model inputs is critical to model accuracy. This paper focuses on inputs screening for hot metal silicon content model. Maximal Information Coefficient (MIC) is adopted to analyse associations between blast furnace variables, and five regression algorithms were used to test the modeling accuracy. The simulation tests verified the importance of input screening for modeling.