Monash University
monash_162374.pdf (5.09 MB)

A context-aware traffic congestion estimation framework to overcome missing sensory data in Bangkok

Download (5.09 MB)
posted on 2017-03-01, 01:40 authored by Raphiphan, Panraphee
Traffic congestion is a problematic experience for commuters in metropolitan areas, costing time and money. For situations involving emergency services, traffic congestion may also be life threatening. A traffic information system (TIS) can play a significant role in improving traffic congestion problems by providing information to road users about the location and degree of traffic congestion. Knowing the location and degree of traffic congestion in real time can assist drivers in planning their routes to avoid heavy traffic. A TIS collects data from multiple sources, including sensors. However, data from sensors may become unavailable due to some reasons such as sensor damage or lost communication. In addition, some roads lack sensors. To ensure the availability and continuity of reported traffic information despite the uncertainty of sensor data, an approach that estimates traffic congestion when sensory data is not available is required. In this thesis, we conduct research into this issue through the lens of a design science research methodology. We propose an artefact, the Context-Aware Traffic Congestion Estimation Framework to Overcome Missing Sensory Data (the CATE framework), to address the above issues. Most existing methods estimate traffic congestion using sensors. In contrast, the CATE framework utilizes available external context information to infer the traffic situation. The framework contains several inference models that represent different situations based on the available context. When sensory traffic data is missing, an appropriate model is selected during run time to infer the traffic congestion degree. The models were developed using machine learning algorithms during our research based on traffic data collected in Bangkok. To deal with the possibility of changes to traffic situations that may make predictions less accurate, the CATE framework incorporates a built-in relearning function that can be used to improve the accuracy of models over time. During the test phase of this research, the CATE framework proved feasible and efficient. It inferred the traffic congestion degree with accuracy higher than that of existing methods and within comparable turnaround times. To further improve the initial artefact of the CATE framework, further test was carried out in the form of survey. The survey aims to validate and improve the initial selection of the context information chosen for the CATE framework. The survey collected Bangkok road users’ perceptions of the factors that affect traffic in Bangkok. The evaluation of this phase demonstrated that the final artefact improved from the initial artefact and again performed better than existing methods in terms of accuracy while also reducing the required processing times and costs associated with calculating the traffic congestion degree. The proliferation of social media and mobile devices suggests that these are possible outlets for disseminating traffic reports in the future and so we included questions in our survey to investigate this possibility. We used the results of these questions to create recommendations for the development of TIS and traffic report services. These recommendations – that information regarding journey routes and traffic conditions be accessible via mobile devices and websites to meet the needs of road users, and that social networks be considered alternative sources of potential traffic data – can be used as guidelines to improve existing TIS and traffic information dissemination services in Bangkok. Through the conceptualization and evaluation of our CATE framework, this thesis makes theoretical and practical contributions to the Intelligent Transportation System (ITS) domain. Through the survey based on the perceptions of Bangkok road users and subsequent statistical analysis, the thesis also makes contributions to the development of TIS reporting systems. Although our study was based on Bangkok data, it may be applicable to other cities that share similar road infrastructure and traffic information issues. This research has produced a framework that has the potential to make a positive difference to road users. The results justify continuing research in this area in order to increase the body of scientific knowledge of the ITS domain and to provide practical support to those involved in managing and maintaining TIS.


Campus location


Principal supervisor

Maria Indrawan-Santiago

Year of Award


Department, School or Centre

Information Technology (Monash University Clayton)

Degree Type



Faculty of Information Technology