Novel Hybrid Method for Travel Pattern Recognition Based on Comparison of Origin-Destination Matrices in Terms of Structural Similarity

Origin-destination (OD) matrices provide transportation experts with comprehensive information on the number and distribution of trips. For comparing two OD matrices, it is vital to consider not only the numerical but also the structural differences, including trip distribution priorities and travel patterns in the study region. The mean structural similarity (MSSIM) index, geographical window-based structural similarity index (GSSI), and socioeconomic, land-use, and population structural similarity index (SLPSSI) have been developed for the structural comparison of OD matrices. These measures have undeniable drawbacks that fail to correctly detect differences in travel patterns, therefore, a novel measure is developed in this paper in which geographical, socioeconomic, land-use, and population characteristics are simultaneously considered in a structural similarity index named GSLPSSI for comparison of OD matrices. The proposed measure was evaluated using OD matrices of smartphone Global Positioning System (GPS) data in Tehran metropolitan. Also, the robustness of the proposed measure was verified using sensitivity analysis. GSLPSSI was found to have up to 21%, 15%, and 9% higher accuracy than MSSIM, GSSI, and SLPSSI, respectively, regarding structural similarity calculation. Furthermore, the proposed measure showed 7% higher accuracy than SLPSSI in the structural similarity index of two sparse OD matrices.