Title Applying Machine Learning to Analyze Vehicular Traffic Information Collected by Wireless Sniffing Technique
Year 2019
Degree Master
School National Cheng Kung University Department of Transportation and Communication Management Science
Author Liang, Teng-Jyun
Summary
      In the collection of traffic data, the most common techniques are the use of vehicle detectors, automatic vehicle identification systems, GPS positioning information, etc. However, each technique has their own disadvantages. Therefore,the development of techniques for collecting traffic data has become indispensable.Due to the popularity of mobile devices such as smartphones in recent years, the use of Wi-Fi Internet and Bluetooth (BT) transmission has increased significantly,especially when smartphones or in-vehicle devices broadcast Wi-Fi or BT signals.It’s worth discussing since the variations of signals with time and space produce some implied traffic information.
      In Fan (2017), the three major problems, the lane identification problem (LIP),the transportation mode problem (TMP) and the multiple devices problem (MDP) are proposed and solved preliminary by seven heuristic algorithms in the scenario of freeway tunnels. In this study, with different ITB topologies (type X, type Rectangle and Type Diamond) in the scenario of road sections and collecting data via two communication technologies (BT and Wi-Fi), the classification and clustering accuracy of three problems predicted by three machine learning models, SVM, KNN and Affinity Propagation, will be presented respectively. For the results in this study, in LIP, the performance of BT is better than that of Wi-Fi ; In terms of the topologies, the performances of type X and type Rectangle are similar, and that of type Diamond is irregular. In TMP, type X and type Rectangle with BT data perform the best, and the accuracies of three topologies with BT and Wi-Fi are above 92%. In MDP, the scenario of two cars in tandem with type X and BT performs the best.
Hashtags
View count: 121