Self Organizing Maps for Traffic Prediction

J. Asamer, K. Din, and W. Toplak (Austria)


Self Organizing Map, Kohonen Net, Neural Net, Traffic Analysis, Traffic Prediction


This paper describes an application of Self Organizing Maps (SOM) in the field of traffic analysis with mid term prognosis. Apart from the (classical) model based approach to analyze and predict traffic, the use of a SOM is described in this work. SOM, a certain form of neural networks, have the ability to visualize multidimensional data in a two dimensional way as well as to cluster and classify them. In this work predictions are done by supplying a priori known information (e.g. day of week, vacation time) to a trained SOM and get the prediction of traffic flow for the desired day. Moreover with the SOM the input data can be analyzed to check influences on a certain value (traffic flow).

Important Links:

Go Back