Preliminary Study: A Principal Component Analysis-Based Approach for Link Selection in Traffic Monitoring

Abstract

The complex and densely populated large urban network forces the transportation networks to deal with a variety of problems, including traffic congestion. Some strategy that can help to optimize traffic flow and reduce traffic congestion is through traffic management systems by monitoring the traffic conditions. The Network Fundamental Diagram (NFD) is a tool that traffic engineers can use to monitor traffic conditions and manage urban traffic networks because it provides a visual representation of the underlying network dynamics. In estimating NFD, identifying influential links (with measurements points i.e. detectors) is a very important part, which has been a key issue in preserving as much information as possible. To select a subset of links that can accurately represent the behaviour of the entire network, a principal component analysis (PCA) is adopted. PCA can be used to remove the variables (links) that contribute minimal information and retained just the variables that contribute the most. The experimental flow-occupancy dataset from 58 inductive loop detectors at urban city was tested and compared with an experiment of four selection approaches. The findings revealed that the PCA-based strategy of Experiment 2 gave result which was more favourable than the results obtained by other experiments. It was able to preserve the shape of the full measurements NFD, retain the information needed (network capacity and critical occupancy within allowable limit) as well as proven to have lowest total Root Mean Square Error (RMSE) value, in comparison with other experiments.