The purpose of the present invention is to provide a system and a method to predict an urban traffic by using a multi-variable pattern recognition model to additionally consider environmental variables of a road except for a detection variable of ITS to predict the state of the road by collecting the divided variables influencing a stagnation. The system comprises: an information extracting unit to extract past record information capable of forming an input pattern vector with information corresponding to a cause which generates the stagnation including an accidental situation such as the change of traffic amount in each time, a geographical characteristic of a road, a geographic condition, an accident, and a construction; a model construction unit to perform an ANN education to know a function relationship between a road pattern and the average pass speed after generating the road pattern vector by executing standardization of a data value according to the extracted past record information; and a traffic prediction unit to predict an average speed of a section by using a local ANN corresponding to an assigned cluster of the input pattern vector for the prediction through the same pre-processing as the model construction unit.
展开▼