Impact of traffic incidents on traffic flow in road networks

With the world constantly improving their standard of living, an increase in usage of motor vehicles can be seen [1]. In a land-scarce country like Singapore, an Intelligent Transport Systems (ITS) is crucial in keeping our traffic network safe and reliable. To maximize traffic network’s efficiency precision traffic and control systems are implemented to monitor and manage traffic flow [2]. With more drivers and vehicles on the road, monitoring traffic flow will allow better prediction of travelling time needed for a motorist to arrive at their destination. Hence, this report gives an overview of my Final Year Project (FYP) on prediction of traffic flow in road networks. The main research focus of the present study is on the efficiency of the traffic prediction models. The road segments are clustered based on their average speed so that each cluster has a similar speed profile. Various number of clusters were analysed to segregate road segments optimally. The prediction model was implemented by Long Short-Term Memory (LSTM) network to capture the auto regressive nature of the time series data. The next aim was to investigate whether incorporating the past speed data of the neighboring road segments would help in capturing the spatial dependencies of traffic speed. For this purpose, past speed features of a sub-network were incorporated in a Support Vectors Machine (SVM) based model. With the integration of spatial-temporal parameters, a significant improvement can be seeing between 38% to 78%. Comparing both model, LSTM has exhibited better performance.Bachelor of Engineering (Electrical and Electronic Engineering