Uncorrected Author Proof Journal of Intelligent & Fuzzy Systems xx (20xx) x–xx DOI:10.3233/JIFS-179644 IOS Press 1 GPS trajectory clustering method for decision making on intelligent transportation systems 1 2 3 Gary Reyes-Zambrano a , Laura Lanzarini b , Waldo Hasperu´ e b and Aurelio F. Bariviera c,∗ 4 a Universidad de Guayaquil, Facultad de Ciencias F´ ı-sicas y Matem ´ aticas. ECG 352, 11000 Guayaquil, Ecuador 5 b Universidad Nacional de La Plata, Facultad de Inform´ atica, Instituto de Investigaci´ on en Inform´ atica LIDI (Centro CICPBA) 1900 La Plata, Buenos Aires, Argentina 6 7 c Universitat Rovira i Virgili, Department of Business, Av. Universitat 43204 Reus, Spain 8 Abstract. Technological progress facilitates recording and collecting information on vehicles’ GPS trajectories on public roads. The intelligent analysis of this data leads to the identification of extremely useful patterns when making decisions in situations related to urbanism, traffic and road congestion, among others. This article presents a GPS trajectory clustering method that uses angular information to segment the trajectories and a similarity function guided by a pivot. In order to initialize the process, it is proposed to segment the region to be analyzed in a uniform way forming a grid. The obtained results after applying the proposed method on a real trajectories database are satisfactory and show significant improvement in comparison with the methods published in the bibliography. 9 10 11 12 13 14 15 Keywords: Segmentation, clustering, GPS trajectories, intelligent transportation systems 16 1. Introduction 17 The growing use of GPS devices and the evolu- 18 tion in the transportation field, demand increasingly 19 efficient techniques for data analysis and decision- 20 making. Intelligent transportation systems process 21 large amounts of GPS trajectory data generated from 22 vehicles on the roads in real time [6, 12, 13]. The 23 collected data must be analyzed to convert them into 24 knowledge in order to use them as support data in 25 decision-making. The detection of traffic congestion, 26 anomalous patterns in traffic that help to predict acci- 27 dents and the evaluation of the performance of main 28 roads and avenues are some of the main application 29 scenarios. 30 ∗ Corresponding author. Aurelio F. Bariviera, Universitat Rovira i Virgili, Department of Business, Av. Universitat 43204 Reus, Spain. E-mail: aurelio.fernandez@urv.cat. As part of data processing, intelligent transporta- 31 tion systems use different algorithms to group GPS 32 trajectories based on different criteria [4, 8, 17]. 33 The bibliography discusses several methods that per- 34 form clustering based on data segmentation and 35 the similarity calculation of these segments. There 36 are different approaches to evaluate the similarity 37 between segments of trajectories according to the 38 type of object and context considered. In the case of 39 GPS trajectories, the function must take into account 40 the underlying graph of the road network and the 41 graphs connectivity or compliance with the sequence 42 order [2]. 43 Among the most used similarity functions in the 44 literature [1] are network-limited distance and dis- 45 tances based on shape and warping. For the purposes 46 of this paper, shape-based distance measurements are 47 of interest as they seek to identify the geometric char- 48 acteristics of trajectories by emphasizing their shape. 49 ISSN 1064-1246/20/$35.00 © 2020 – IOS Press and the authors. All rights reserved