A Hybrid of Computational Intelligence Techniques for Shape Analysis of Traffic Flow Curves
This paper highlights and validates the use of shape analysis using Mathematical Morphology tools as a means to develop meaningful clustering of historical data. Furthermore, through clustering more appropriate grouping can be accomplished that can result in the better parameterizations or estimation of models. This results in more effective prediction model development. Hence, in an effort to highlight this within the research herein, a Back-Propagation Neural Network (BPNN) is used to validate the classification achieved through the employment of Mathematical Morphology (MM) tools. Specifically, the Granulometric Size Distribution (GSD) is used to achieve clustering of daily traffic flow patterns based solely on their shape. To ascertain the significance of shape in traffic analysis, a comparative classification analysis of original data and GSD transformed data is carried out. The results demonstrate the significance of functional shape in traffic analysis. In addition, the results validate the need for clustering prior to prediction. It is determined that a span of two through four years of traffic data is found sufficient for training to produce satisfactory BPNN performance.
W. Kayani et al., "A Hybrid of Computational Intelligence Techniques for Shape Analysis of Traffic Flow Curves," TRB 94th Annual Meeting Compendium of Papers, Transportation Research Board, Jan 2015.
Transportation Research Board 94th Annual Meeting (2015: Jan. 11-15, Washington, DC)
Engineering Management and Systems Engineering
Electrical and Computer Engineering
Center for High Performance Computing Research
Keywords and Phrases
Backpropagation; Neural networks; Traffic data; Traffic flow; Validation
Article - Conference proceedings
© 2015 Transportation Research Board, All rights reserved.