Fuzzy Learning Classifier System and its Application Research in Automatic Traffic Incident Detection

According to the application of pattern classification, a fuzzy learning classifier system was described in detail, including its framework and each important module. Furthermore, the paper also designed the classifier encoding, fitness parameter definition, etc. Then this system was applied in automatic traffic incident detection using I-880 field data and compared with BP network on the detection performance. The BP network was improved with Levenberg-Marquardt method. The comparison result showed that the fuzzy learning classifier system designed in this paper had lower false detection rate and shorter detection time, which fit for the further field application.