We propose three designs for clustering objects: a new graph partitioning algorithm, Boruvka’s algorithm, and a randomized algorithm for object graph clustering. Several points are innovative in our approach to clustering: (1) the randomized algorithm represents a new approach to the problem of clustering and is based on probabilistic combinatorics. (2) All of our algorithms can be used to cluster objects with multiple connectivity. (3) Currently applied partition-based clustering algorithms are based on Kruskal’s algorithm which always runs significantly slower than Prim’s and also uses considerably more storage. However in our implementation of clustering algorithms we achieved additional reduction in processing time of object graphs.