Recognition of urban road traffic pattern is an important part of intelligent transportation systems .An enormous number of traffic data could be obtained with the develop ment of in formation techniques . This motivates the application of machine learn ing in the road traffic area, especially in traffic incident detection. Incident detection algorith m in the machine learning can be defined as a binary classification problem, where each occurrence is the traffic state on a road segment at a particular time. This paper is concerned with how to detect traffic anomaly patterns in an urban road network by using potential sensor data. In this paper, by using Simulation of Urban Mobility (SUM O) software, we have chosen to work on the Chula -Sathorn SUM O Simulator (Chula -SSS) dataset. SUMO enables users to simulate traffic networks and supports the traffic data by setting up conveniently simu lated lane area detectors. By using calibrated Chula -SSS dataset, anomaly traffic patterns have been generated and classified with the support vector mach ine algorithm with the radial basis function. The algorith m has been shown here to detect accurately of at least 87% (and 71 %) of the simulated lane-closure incidences, by rely ing on sensors fro m (i) within the incident area, and (ii) at the upstream as well as downstream areas adjacent to that incident link, respectively.