It is significant for the GPS positioning system to control the vehicle and analyzes the congestion. But there are bad values in GPS sampling data, easy to produce large error on the analysis results, thus affecting the traffic management decision. This paper completes the original data segment, field extraction and bad values cleaning using Gauss mixture model, K- means clustering analysis and SOM self-organizing neural network separately. Thees three methods are mainly used for data clustering analysis, cleaning the isolated points according to the results. The results show that the recognition accuracy of Gauss clustering and K- clustering algorithm is less than SOM self-organizing neural network, but the operating efficiency of the first two algorithms is better than the latter.