Open AccessINFORMATION TECHNOLOGY FOR TRAJECTORY DATA MINING
Author(s) -
Maryna Sydorova,
O. H. Baybuz,
Olha Verba,
Pavlo Pidhornyi
Publication year - 2021
Publication title -
science and innovation
Language(s) - English
Resource type - Journals
eISSN - 2413-4996
pISSN - 2409-9066
DOI - 10.15407/scine17.03.078
Subject(s) - computer science , cluster analysis , data mining , trajectory , software , object (grammar) , visualization , motion (physics) , hierarchical clustering , similarity (geometry) , artificial intelligence , image (mathematics) , physics , astronomy , programming language
. Advanced technologies allow almost continuous tracking and recording the movement of objects inspace and time. Detecting interesting patterns in these data, popular routes, habits, and anomalies in object motion and understanding mobility behaviors are actual tasks in different application areas such as marketing, urban planning, transportation, biology, ecology, etc.Problem Statement. In order to obtain useful information from trajectories of moving objects, it is important to develop and to improve mathematical methods of spatiotemporal analysis and to implement them in highquality modern software.Purpose. The purpose of this research is the development of information technology for trajectory data mining.Materials and Methods. Information technology contains the three main algorithms: revealing key pointsand sequences of interest with the use of density-based trajectories clustering of studied objects; detecting patterns of an object movement based on association rules and hierarchical cluster analysis of its motion trajectories in the time interval of observations, similarity measure of the motion trajectories has been proposed to be calculated on the basis of the DTW method with the use of the modified Haversine formula; new algorithm for revealing permanent routes and detecting groups of similar objects has been developed on the basis of clustering ensemblesof all studied trajectories in time. The clustering parameters are selected with multi-criteria quality evaluation.Results. The modern software that implements the proposed algorithms and provides a convenient interactionwith users and a variety of visualization tools has been created. The developed algorithms and software have beentested in detail on the artificial trajectories of moving objects and applied to analysis of real open databases.Conclusions. The experiments have confirmed the efficiency of the proposed information technology thatmay have a practicable application to trajectory data mining in various fields.