Abstract: A new heuristic density-based ant colony clustering algorithm (HDACC) is presented.The device of memory bank is first proposed, which brings forth the heuristic knowledge guiding ant to move in the bi-dimensional grid space. In this way, the algorithm's convergence is speeded up and the appearance of "un-assigned data object" avoided, and the error rate in classification drops subsequently. A density-based method is then proposed permiting each ant to look ahead and reduces the number of times in region-inquiry. Consequently the clustering time is saved. Some experiments were made on real and synthetic date sets. Experimental results were compared with those obtained using other classical clustering algorithms. The results demonstrated that the proposed HDBCSI is a viable and effective clustering algorithm.