This paper describes the pattern-based forecasting method that was developed in the European Communities' DRIVE KITS (Knowledge Based Intelligent Traffic Control Systems) project. The prototype of the forecasting method is structured in three main parts. These are: (1) the pattern transformation; (2) the pattern classification; and (3) the choice of a suitable comparison pattern including the forecasting. Additionally, there are two knowledge-bases. The first one contains the basic knowledge for pattern classification using week-day groups and additional site-specific information. The second base is filled with acquired and transformed typical patterns. The forecasting method is based on both traffic volumes measured at a specific location and on classification knowledge. The method was tested on the KITS field trial in the German city of Köln. It is concluded that the developed pattern-based forecasting method has several advantages compared to conventional approaches. Important advantages are the openness and the explanation possibilities.
Samenvatting