Metro Travel Pattern Identification: Cluster Analysis Based on Daily OD Matrix

It is essential for metro operations to identify typical daily OD matrix from the long-term automatic fare collection (AFC) data. A daily metro travel demand pattern identification method is proposed, using eight-month AFC data from Shanghai metro networks in China. For the high dimension problem of the long-term OD matrix, principal component analysis (PCA) was applied to reduce the dimension of the OD matrix. Then, fuzzy C-means clustering was used to identify the daily metro travel demand patterns. The results show metro operating days can be clustered into three typical daily travel categories, namely weekdays, weekends, and holidays. The metro OD travel flow data processing framework proposed in this study identifies the typical operating daily travel patterns of complex networks, integrating dimensionality reduction technology and clustering methods, which provides a new solution for understanding the daily mobility structure, which is helpful for short-term travel demand forecasting and metro operation schedule.