Ελληνικό Μεσογειακό Πανεπιστήμιο
Event prediction using spatial-temporal data: predictive traffic accident approach through categorical logic
Abstract
dc:descriptionMany scientific fields show great interest in the extraction and processing of spatiotemporal data, such as medicine with an emphasis on epidemiology and neurology, geology, social sciences, meteorology, and a great interest is also observed in the study of transport. Spatiotemporal data differ significantly from spatial data, since spatiotemporal data refer to measurements, which take into account both the place and the time in which they are received, with their respective characteristics, while spatial data refer to and describe information related only to place. The innovation brought about by spatiotemporal data mining has caused a revolution in many scientific fields, and this is because through it we can now provide solutions and answers to complex problems, as well as provide useful and valuable predictions, through predictive learning. However, combining time and place in data mining presents significant challenges and difficulties that must be overcome. Spatiotemporal data mining and analysis is a relatively new approach to data mining which has been studied more systematically in the last decade. The purpose of this dissertation is to provide a good introduction to spatiotemporal data, and through this detailed description, we attempt to introduce description logic and gain a complete knowledge of these data. We aim to introduce a new way of describing them, aiming for future studies, by combining the expressions that arise by type of data, using description logic, with new expressions, that can be derived, to describe future states of objects and environments with great precision, providing accurate predictions. In order to highlight the value of spatiotemporal data, we proceed to give a brief description of ST data in the introduction, but we also make special reference to the prediction of events and more specifically traffic accidents. We describe the relevant work carried out to date, the types of spatiotemporal (ST) data, their properties and the transformations that can be made between them, attempting, to a small extent, to introduce constraints and rules using description logic, introducing description logic into spatiotemporal data by type, when initially presenting the ST data. The data snapshots by species and similarities between the cases are then described. We present methods, introducing clustering, dynamic ST clusters, predictive learning, pattern mining frequency, and pattern emergence, and problems such as anomaly detection, identifying time points of changes in the behavior of the observed object, and development of relationships between them. We describe the application of ST data in various fields today, as well as the future work. The goal is to export our conclusions, about the representation and study of spatiotemporal data can, in combination with other properties which accompany all natural phenomena, through their appropriate processing, lead to safe conclusions regarding the study of problems, and also with great precision in the extraction of predictions by accurately determining future states of an environment or an object. Thus, the importance of ST data, especially ST data of type events, makes them particularly valuable today in various scientific fields, and their extraction is a particularly demanding challenge for the future.Μore specifically, an event that takes place at a specific time and location is an event, which can be either weather-related (snowfall), social (crime), natural (earthquake), political (political unrest), or even medical in nature (pandemic). They are events that do not belong to the spectrum of "normal" or "usual" events and come to change a given situation. Their prediction would be very beneficial, both in their immediate and timely response and in their avoidance, especially when we refer to events with negative impacts on humans, for example avoiding a traffic accident, through prevention. This is currently a challenge for researchers, who are called upon to manage and analyze a huge volume of data, in order to design applications that will lead to safe prediction of events, having as resources the use of artificial intelligence and high computing power. Nevertheless, great progress has been made in this direction, however, the heterogeneity of the input data that a forecasting application is called upon to process, in terms of their nature (spatial, temporal and semantic), with the corresponding complex dependencies between them, constitutes today the greatest challenge for researchers. For this reason, the first forecasting applications deal with the processing of data that come from specific, in terms of their number and characteristics, situations, while at the same time having the possibility that these applications can respond to different situations e.g. an application that deals with the prediction of a pandemic, can also cope with the prediction of a weather phenomenon, simply by introducing a different type of data. In this dissertation, we will present the forecasting applications that have been designed to date. Mainly, we will deal with the study of traffic accident prediction. We will present a prediction model, using description logic, creating a Knowledge Base, using the Analysis algorithm as a proof method, we will study and analyze all possible scenarios with different conditions that arise each time. More specifically, we implement a map-based traffic accident prediction model, in which the user will define a route and values will be extracted, such as the coordinates of his location, the time he wishes to drive, the prevailing weather conditions and whether the route contains dangerous points, such as sharp turns, as well as the driver's sobriety, i.e. whether the driver has consumed alcohol or not. These values will be entered and updated in a knowledge base in a prolog file, every 5 seconds, and then, based on the content of the knowledge base, a jpl file will be updated, extracting conclusions using and executing the required queries. The model also allows the user to send a message (email), with the current coordinates of the vehicle (auto-completed), to a roadside assistance company for immediate assistance, in case of an accident. The most important advantages of our model are that does not require much computing power or a large volume of data (big data) and yields particularly high success rates.
Degree
thesis:*- Grantor dc:publisher
- Ελληνικό Μεσογειακό Πανεπιστήμιο
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Koutsaki, Eleftheria
- Κουτσάκη, Ελευθερία
Subjects
dc:subject × 36- Συμβάντα
- Πρόβλεψη συμβάντων
- Τροχαίο ατύχημα
- Μεγάλα δεδομένα
- Χωροχρονικά δεδομένα
- Τεχνητή νοημοσύνη
- Κατηγορική λογική
- Βάση γνώσεων
- Αλγόριθμος ανάλυσης
- Γλώσσα prolog
- Αρχείο JPL
- Spatiotemporal data
- Data mining
- Geographical Data
- Time series
- Raster
- Point
- Events
- Event prediction
- Traffic accident
- Artificial intelligence
- Description logics
- Knowledge base
- Resolution algorithm
- PROLOG
- JPL
- OSRM
- Spatial Temporal
- Επιστήμη Ηλεκτρολόγου Μηχανικού, Ηλεκτρονικού Μηχανικού, Μηχανικού Η/Υ
- Επιστήμες Μηχανικού και Τεχνολογία
- Αυτοματισμοί και Συστήματα ελέγχου
- Ρομποτική
- Electrical Engineering, Electronic Engineering, Information Engineering
- Engineering and Technology
- Automation and control systems
- Robotics
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/59311