Blast furnace gas utilization rate is one of the indicators for measuring the smooth operation of the blast furnace. The prediction model of the blast furnace gas utilization rate based on the extreme learning machine algorithm (ELM) is firstly established. The burden surface characteristics and the indexes of the blast furnace condition are the input parameters, and the blast furnace gas utilization rate is the output parameter. In most cases, the regular item factor is introduced for ELM to ensure satisfactory output. In this paper, the same prediction model based on PCA-ELM algorithm which is based on the principal component analysis method (PCA) and ELM is established secondly. Real production data of the blast furnace is used to verify the prediction model. By comparing with the results of two models, the model based on the PCA-ELM algorithm has better accuracy than that based on ELM.