A Study on Travel Patterns in Kancheepuram City Using Unsupervised and Fuzzy Clustering Techniques

Abstract: Urban mobility and spatial development are two critical dimensions of urban planning that profoundly shape the efficiency of transportation networks, accessibility of essential services, and the overall quality of life for city residents. Urban mobility refers to the ease and effectiveness with which people move within a metropolitan area. At the same time, spatial development pertains to the spatial distribution of infrastructure, buildings, and services that enable this movement. These elements are interdependent and have a cascading effect on cities' socioeconomic and environmental sustainability. In this context, analyzing travel patterns becomes vital for optimizing urban transportation systems, particularly in regions experiencing rapid urbanization and evolving mobility needs.
One of the most fundamental datasets for understanding urban travel behaviour is Origin-Destination (O-D) data, which captures the starting and ending points of individual trips. O-D data serves as a cornerstone for analysing the demand and flow of traffic within a city, offering critical insights into how people and vehicles move across different urban sectors. This dataset allows urban planners to identify key travel corridors, peak traffic zones, and the influence of various socio-economic and environmental factors on travel demand. Such analyses support the design of more efficient transportation systems, balanced land-use policies, and strategies to improve mobility across the urban landscape.
This study centers on the application of machine learning techniques to analyse and predict urban travel patterns in Kancheepuram, a district that exemplifies the challenges of urban mobility in developing urban regions. Specifically, the study employs clustering algorithms to process and interpret O-D data, to uncover hidden patterns in travel behaviour. The analysis uses clustering algorithms such as k-means, hierarchical clustering, and fuzzy c-means to group travel patterns based on vehicle type, trip purpose, and occupancy levels. These clustering methods provide a nuanced understanding of travel behaviours, including trip distances, peak travel times, and frequently used travel routes. The performance of each clustering algorithm was rigorously evaluated using silhouette and Dunn’s indices which measure the compactness and separation of clusters.
The results of this work revealed distinct travel patterns that offer new insights into urban mobility in Kancheepuram. These findings have significant implications for urban transportation planning, providing a more data-driven approach to resource allocation, congestion management, and infrastructure development. The major takeaway is the integration of pattern-finding algorithms with traditional transportation analysis techniques to forecast future travel demand more accurately, allowing city planners to proactively design more resilient and efficient transportation systems. These findings are not only relevant to Kancheepuram but also befit broader applicability for cities facing similar challenges in managing urban growth and transportation demand.