Daily metro origin-destination pattern recognition using dimensionality reduction and clustering methods

Abstract

Widespread usage of Smart Cards is leading to unprecedentedly massive growth of the quantity of data. However, traditional methods still fail to fully recognize mobility patterns and a better way of data mining is to be explored. In order to achieve reliable pattern recognition results, principal component analysis and singular value decomposition are respectively applied. Based on the dimensionality reduced matrix, affinity propagation is selected as a suitable clustering algorithm to recognize demand patterns. Spectral clustering is introduced to make a comparison. Different clustering evaluation indicators are used to serve as objective references. Representative categories are clustered, which correspond to weekdays, weekends, holidays, and different months, respectively. The integration of dimensionality reduction and clustering offers a new way to understand daily mobility structure. To metro system operators, this study also provides information on traffic volume variation and temporal distribution of the whole year. Besides, the procedures of dealing with daily demand matrix can be applied in traffic planning, management and operation.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 21 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.