Abstract: In order to more accurately predict the stability of tailings dam, the training speed and simple parameters and high accuracy of extreme learning machine (ELM) model, in view of the ELM model randomly generated in the process of training the connection weights and threshold, the hidden layer in generalization ability insufficiency, the model of stability problems, introduced based on linear weighting method improved particle swarm optimization (IPSO) for its optimization, predicting the stability of tailings dam is proposed to improve the particle swarm optimization extreme learning machine (IPSO-ELM) model.To apply the model prediction in tailings dam instance, in the selection of 35 sample data, the first 30 group as the training sample, after 5 groups as the test sample, friction angle, slope angle, tailings dam materials within severe, pore pressure ratio, cohesion and the slope height 6 tailings dam stability influence factors as input parameters, in tailings dam stability safety factor for the output parameters, the prediction results and the ELM model and PSO-ELM model comparison, the results show that IPSO-ELM model has higher prediction precision, close to the actual and estimated values, the reliability and effectiveness of IPSO-ELM model in tailing dam stability evaluation are verified.