The present invention provides a bee-swarm optimization method and device. The bee-swarm optimization method comprises the following steps of: according to a given preset data set, establishing an undirected graph based on mutual information; processing the undirected graph to a communication undirected graph; performing primary optimization of the communication undirected graph, eliminating extraedges in the communication undirected graph, and obtaining a first optimization graph; employing the bee-swarm algorithm to perform secondary optimization of the first optimization graph, supplementing edges deleted in the first optimization graph, determining the direction of each edge, and obtaining a second optimization graph; and employing the second optimization graph as an initial search graph of the bee-swarm algorithm to perform search of an optimal value. The bee-swarm optimization method employs the complete second optimization graph with directions as the initial search graph of the bee-swarm algorithm to effectively reduce the probability of the bee-swarm algorithm being got in the local optimization and allow the bee-swarm optimization to obtain an optimal search result as far as possible.