The fast-growing large point of sale databases in stores and companies sets a pressing need for extracting high-level knowledge. Transaction clustering arises to receive attentions in recent years. However, traditional clustering techniques are not useful to solve this problem. Transaction data sets are different from the traditional data sets in their high dimensionality, sparsity and a large number of outliers. In this paper we present and experimentally evaluate a new efficient transaction clustering technique based on cluster of buyers called caucus that can be effectively used for identification of center of cluster. Experiments on real and synthetic data sets indicate that compare to prior work, caucus-based method can derive clusters of better quality as well as reduce the execution time considerably.