Abstract: In order to investigate the effects of steel fiber types and dosage on the mechanical properties of basalt fiber reinforced concrete（BFRC）, three types of steel fibers（end-hook type, wave type, and copper-plated type） were added to BFRC at steel fiber volume dosage of 0, 0.8%, 1.0%, and 1.2%, respectively. Grey theory GM（1,9） and BP neural network models were established to predict the compressive strength of concrete and their reliabilities were compared, and nuclear magnetic resonance testing was combined to test the change patterns of pores in concrete. The results show that with the increase of steel fiber content, the compressive strength of BFRC first increases and then decreases. Among them, the group with the best mechanical properties is the end-hook type steel fiber group with a content of 1.0%. The compressive strength at 14 days is 44.51 MPa. Due to surface corrosion of the steel fiber, the compressive strength at 28 days decreases to 43.13 MPa. The grey theory GM（1,9） model and the BP neural network prediction model are established, for which the prediction accuracy of both models meets the requirements. However, the BP neural network has higher prediction accuracy and better stability, with an average relative error of 3.81%, and the average relative error of the grey theory model is 7.55%. As the volume of steel fiber increases, the proportion of cementitious pores and capillary pores in concrete first increases and then decreases. Therefore, reasonable addition of steel fibers can optimize the pore structure in BFRC and improve its concrete compressive strength.