Home // INTELLI 2020, The Ninth International Conference on Intelligent Systems and Applications // View article

Clustering Techniques for On-Demand Transport Data: A Case Study

Authors:
Carlos Afonso
Ana Alves

Keywords: On-Demand Transport; Transport Requests; Partition-based Clustering; Density-based Clustering; K-Means; DBSCAN

Abstract:
The on-demand transportation request requires a quick and efficient response to satisfy customers and also make the system a viable option. Clustering techniques were used to group transport requests, i.e., the starting points of vehicles that have been requested to optimize the service with the benefit of reducing the number of vehicles needed and, consequently reduce the amount of pollution produced. The objective is to compare the two main clustering techniques from two distinct categories: partitioned and density-based to evaluate which one is best suited for defining start zones. The quality of the generated clusters is defined by calculating the silhouette related to the generated clusters. Using previous references, the two clustering methods were compared based on the desired characteristics. The analysis demonstrates that DBSCAN is best suited for the problem and is then applied over a sample dataset. The manner in which the DBSCAN algorithm can generate random shapes, which fit well into the geographic distribution of points and how the number of necessary clusters do not need to be defined in advance makes it the ideal choice for defining starting zones.

Pages: 12 to 17

Copyright: Copyright (c) IARIA, 2020

Publication date: October 18, 2020

Published in: conference

ISSN: 2308-4065

ISBN: 978-1-61208-798-6

Location: Porto, Portugal

Dates: from October 18, 2020 to October 22, 2020