  Citation: Erdeli´ c, M.; Cari´ c, T.; Erdeli´ c, T.; Tišljari´ c, L. Transition State Matrices Approach for Trajectory Segmentation Based on Transport Mode Change Criteria. Sustainability 2022, 14, 2756. https: //doi.org/10.3390/su14052756 Academic Editor: Sadko Mandzuka and Kresimir Vidovic Received: 20 December 2021 Accepted: 23 February 2022 Published: 26 February 2022 Publisher’s Note: MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affil- iations. Copyright: © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/). sustainability Article Transition State Matrices Approach for Trajectory Segmentation Based on Transport Mode Change Criteria Martina Erdeli´ c* , Tonˇ ci Cari´ c , Tomislav Erdeli´ c and Leo Tišljari´ c Faculty of Transport and Traffic Sciences, University of Zagreb, Vukeli´ ceva 4, 10000 Zagreb, Croatia; tcaric@fpz.unizg.hr (T.C.); terdelic@fpz.unizg.hr (T.E.); ltisljaric@fpz.unizg.hr (L.T.) * Correspondence: merdelic@fpz.unizg.hr Abstract: Identifying distribution of users’ mobility is an essential part of transport planning and traffic demand estimation. With the increase in the usage of mobile devices, they have become a valuable source of traffic mobility data. Raw data contain only specific traffic information, such as position. To extract additional information such as transport mode, collected data need to be further processed. Trajectory needs to be divided into several meaningful consecutive segments according to some criteria to determine transport mode change point. Existing algorithms for trajectory segmentation based on the transport mode change most often use predefined knowledge- based rules to create trajectory segments, i.e., rules based on defined maximum pedestrian speed or the detection of pedestrian segment between two consecutive transport modes. This paper aims to develop a method that segments trajectory based on the transport mode change in real time without preassumed rules. Instead of rules, transition patterns are detected during the transition from one transport mode to another. Transition State Matrices (TSM) were used to automatically detect the transport mode change point in the trajectory. The developed method is based on the sensor data collected from mobile devices. After testing and validating the method, an overall accuracy of 98% and 96%, respectively, was achieved. As higher accuracy of trajectory segmentation means better and more homogeneous data, applying this method during the data collection adds additional value to the data. Keywords: urban mobility; transport mode; real time change point detection; mobile phone sensor data; transition state matrices 1. Introduction In many traffic management applications, data collected through sensor technologies provide a means for estimating traffic demand and transport planning. Recognition of daily transport activities has several applications, such as educating users with the aim of changing their behavior patterns or to encourage healthier lifestyle. In addition, applica- tions that track user mobility are helping to plan public urban transportation, track vehicle traffic, and support smart parking [1]. Data collected from mobile devices are time series of sensor values recorded during user movement through the traffic network. The collected data do not contain additional information such as transport mode, so further processing has to be applied to extract such data from the collected data set. With the increase in machine learning methods, various methods are applied on sensor-based data to obtain additional information such as human activities [2], transport modes [3] or congestion zones [4]. The additional information added to a large amount of data provide a base for developing predictive models in the field of urban mobility that capture hidden characteristics of traffic flow [5], user behavior [6], or interactions of transport network users [7]. One of the basic processing steps of traffic mobility data is trajectory segmentation, where a raw trajectory is divided into several meaningful consecutive parts. A method for Sustainability 2022, 14, 2756. https://doi.org/10.3390/su14052756 https://www.mdpi.com/journal/sustainability