Rough set theory is an effective method for dealing with uncertain boundary objects. The rough K-means clustering algorithm which combines rough set with K-means is simple and efficient. Though it can deal with clustering boundary elements, it has some drawbacks, for instance, the original rough K-means clustering algorithm is sensitive to the initial center, the set-up of empirical weigh ignores data difference, the unreasonable threshold setting engenders fluctuation of clustering results. To tackle these drawbacks, this paper proposed an improved rough K-means clustering algorithm combined with ant colony algorithm. The improved algorithm is optimized for rough K-means clustering by using random probability selection strategy and pheromone update of positive and negative feedback mechanisms in ant colony algorithm, and using dynamic threshold adjustment algorithm and associated weights method. Finally, the UCI’s Iris set, Balance-scale set and Wine set are used for verification of the algorithm. The results show that this algorithm exhibits a higher clustering accuracy.