Abstract: Cluster analysis is an important branch of non-supervised model classification, DBSCAN (Density-Based Spatial Clustering of Applications with Noise) is one of the most common algorithms in density-based clustering methods. It's widely researched and applied in many fields as it can find clusters of arbitrary shapes with noises. Some shortcomings of DBSCAN and also recently improved algorithms based on DBSCAN are focused on. A new data partitioning method is proposed to solve the problem that m_pts-HDBSCAN clustering quality will degrade when applied in varied density dataset. Firstly the proposed partitioning method calculates the numbers of the group based on the histogram of the data distribution. Secondly it is determined whether to partition the dataset based on the threshold value. Sub-datasets generated by partitioning method will bind with m_pts-HDBSCAN to find clusters and finally merge the sub-clusters to one. Experiment shows the proposed binding algorithm is more effective than m_pts-HDBSCAN in finding clusters when dataset density is not even.