The accurate prediction of steel prices is helpful for construction companies to formulate reasonable material procurement strategies. For the current steel price prediction research, the long memory of its price changes is not considered, resulting in the loss of effective information in the modeling process and the increase of prediction errors. In this paper, the ARFIMA steel price prediction model considering long memory was established. The steel price prediction was carried out based on the price of rebar in Qingdao from January 2014 to June 2019. The predicted values of the ARFIMA model and ARIMA model were used for comparative analysis of the true value. The experimental results show that the accuracy of the steel price prediction of the ARFIMA model is 1.7% higher than that of the ARIMA model, and the prediction effect is more stable.