With the rapid development of computer technology and information technology, the development of blockiness prediction methods has changed from traditional empirical formulas and statistical models to numerical simulation and artificial intelligence technology. In order to further improve the production efficiency of open-air step blasting in highway slope construction, the prediction accuracy of large block rate was improved. In this paper, a blasting effect prediction model is proposed. Eight parameters, such as step height, step slope angle, and minimum resistance line, were selected as input variables. An improved tuna swarm optimization (TSO) burst prediction model based on Extreme Learning Machines (ELM) was constructed to predict the burst ratio. The fitness curve and prediction error were used as evaluation indexes to confirm the effectiveness and practicability of the prediction method adopted in this project.