Recently from October, 2016 until April, 2017 in Seoul, South Korea, a series of protests have occurred on almost every Saturdays, where crowds estimated from 50,000 to 2.3 million have gathered in streets around palatial Gwanghwamun Square due to a political scandal of former president Park Geun-hye. In total, nearly 16.5 million crowds have attended to these events by holding candles aloft with wishes for an immediate resignation of the former president. With respect to these consecutive events in South Korea, we present a research with a goal of finding spatiotemporally the most influential factors for the characterization of event size (i.e. the number of participants) by using Seoul taxi trajectory data. For this objective, an analysis was conducted that finds the optimal combination of variables which maximizes dissimilarity measures among event size categories by utilizing Dunn's Index (DI) as an evaluation function, for which Genetic Algorithm (GA) was used due to combinatorial computation complexity of the optimization problem. As a result, the most governing regions and time bins with respect to three taxi statuses: drop off, pick up, passing with and without passenger were found for event size categories. Based on the analysis, we could come up with a reverse engineering approach which can find a list of influential factors of taxi trajectory data for characterizing event size. The influential factors could be used for traffic control and transit service plans for city administrators. The proposed approach can be applied to any other set of given clusters having many variables that requires huge amount of combinations to pick governing factors for such characterization problems.
|Title of host publication||Proceedings of 1st ACM SIGSPATIAL Workshop on Analytics for Local Events and News, LENS 2017|
|Publisher||Association for Computing Machinery, Inc|
|Publication status||Published - 2017 Nov 7|
|Event||1st ACM SIGSPATIAL Workshop on Analytics for Local Events and News, LENS 2017 - Redondo Beach, United States|
Duration: 2017 Nov 7 → …
|Name||Proceedings of 1st ACM SIGSPATIAL Workshop on Analytics for Local Events and News, LENS 2017|
|Other||1st ACM SIGSPATIAL Workshop on Analytics for Local Events and News, LENS 2017|
|Period||17/11/7 → …|
Bibliographical notePublisher Copyright:
© 2017 ACM.
All Science Journal Classification (ASJC) codes
- Computer Vision and Pattern Recognition
- Computer Graphics and Computer-Aided Design