Abstract: In the context of intelligent transportation systems, millimeter-wave radar provides real-time point cloud data. The ultimate goal of processing millimeter-wave point clouds is to extract and classify targets. However, current clustering algorithms only focus on target extraction and do not handle classification, thus the quality of clustering needs to be considered and improved for the accuracy of final classification. To address this issue, this paper proposes a joint iterative clustering algorithm for millimeter-wave radar traffic target classification. It initially performs clustering using a fast density clustering algorithm based on constructing k-d tree, followed by designing a machine learning classification algorithm based on kernel function SVM to classify the clustering results. In cases where clusters are classified as abnormal data, this paper iteratively adjusts the clustering parameters. This method optimizes between clustering and classification iteratively, improving the accuracy of both simultaneously. Experimental results on millimeter-wave radar data demonstrate that the proposed method increases the accuracy from 76.77% to 96.46% compared to non-iterative algorithms.