Abstract: A radial basis function neural network model combined with particle swarm optimization algorithm and independent component analysis is proposed to predict the endpoint of BOF(basic oxygen furnace)steelmaking.In order to solve the issues that the objective function falls into the local optimum and the sequence of independent components is uncertain,this paper utilizes the global ergodicity of particle swarm optimization algorithm and the local optimizing capacity of fast fixed-point algorithm to improve the traditional independent component analysis algorithm,as well as the redundant information is compressed and the input dimension is reduced.The extracted independent features are introduced into the radial basis function neural network to predict the endpoint temperature and carbon content.Simulations are made with the practical data of BOF production,and the result proves the proposed model can improve the accuracy and reassure the reliability of prediction.