Aiming at the characteristics of the gas outburst disasters, the double chaos search bee colony (DBC) algorithm for gas outburst prediction is proposed. DBC algorithm improves ABC by introducing the chaos optimization idea in the process of population evolutionary. The chaos method generates initial population in order to improve the quality of the food source, to prevent premature convergence of the algorithm. Meanwhile, chaos idea is introduced to the bee swarm local search. Through the method, this drawback of getting in local minima is resolved. Then DBC is merged with MLPNN to construct a gas outburst prediction model by using DBC to train the weights and thresholds. The predicted results show that the method is accurate.