The invention discloses a clustering method and device and belongs to the technical field of network. The method includes: acquiring at least one first feature of a search keyword used by a user; acquiring at least one second feature of the search keyword; acquiring at least one third feature of the search keyword; acquiring at least one clustering feature according to each first feature, each second feature and each third feature; clustering according to each clustering feature. The method has the advantages that by combining and processing various features, the problem of poor clustering results caused by excessively sparse features, noise of clustering results is reduced, and optimal clustering results can be achieved.