A Classification Algorithm For Traffic State Based On Ensemble Fuzzy Classifier

It is important to judge traffic state rapidly and accurately, which is alsoan important research content of Advanced Traveler Information System (ATIS).Releasing the information of road traffic in time can offer the optimal route to drivers soas to avoid traffic jam effectively. Therefore, studying the traffic state distinctionalgorithm has an important theoretical and practical significance. This is a researchcontent of Beijing Natural Science Foundation-funded projects named as "Research onTraffic Flow Model and Key Technologies of Traffic Guidance for Beijing UrbanFreeways" (8052016).Traffic state has fuzziness and subjective characteristic, which is compatible withfuzzy logic. Therefore, fuzzy inference is used in distinguishing traffic state. Also, astraffic is affected by many factors and traffic state is not certain, multiple classifiers byensemble leaming can distinct traffic state comprehensively to enhance the accuracy ofthe classification. Fixed detectors in some cities have achieved seamless cover and canprovide real-time basic traffic information. In view of the fixed detector data, thisdissertation proposes a classification algorithm of traffic state based on ensembleleaming and fuzzy inference to improve the algorithm in common use.Microwaves detector data of Beijing 2^nd ring and 3^rd ring are used to validate thealgorithm. The results by Matlab programming prove that the new algorithm is betterthan existing algorithm and can reflect traffic state accurately. Using Visual C++ anddatabase to realize simulation software, the software can realize algorithm training, thedisplay of road traffic state and other functions. The research ideas and the algorithmproposed in this dissertation is a method exploration of traffic state distinction algorithmof fixed detectors and may provide certain theoretical reference.