{"id":{"repo_id":"uts","oai_identifier":"oai:opus.lib.uts.edu.au:10453/177936"},"canonical_url":"https://search.dev.ndltd.org/etd/uts/oai:opus.lib.uts.edu.au:10453/177936","repository":{"repo_id":"uts","name":"University of Technology Sydney","base_url":"https://opus.lib.uts.edu.au/oai/request"},"display":{"title":"Advanced AI techniques for comprehensive traffic incident analysis: enhancing incident duration prediction and accident risk forecasting","abstract":"The progress of global urbanization and growth of vehicular traffic have led to an increase in traffic incidents, increasing demand for efficient modeling and prediction methodologies essential for traffic management. To meet this challenge, the thesis proposes the application of advanced machine learning and deep learning techniques to enhance traffic incident modeling. The research contributions are outlined as follows: Development of a Universal Framework: A novel framework is introduced for predicting traffic incident duration across various road network layouts, handling classification and regression problems, addressing outliers and imbalanced data classes. This framework is designed to handle both classification and regression problems effectively, while also addressing the challenges of outliers and imbalanced data classes. It utilizes feature importance estimation methods, such as SHAP, to identify critical variables, and incorporates anomaly detection techniques like One-Class SVM and Isolation Forest into the prediction models. Exploration of Data Fusion Techniques: A key contribution of this research is the exploration of data fusion. By integrating different data types, including traffic flow information, textual incident descriptions, and historical traffic flow data, the thesis proposes a methodology to enhance incident duration prediction accuracy. This is achieved through the use of deep learning methodologies like LSTM and ANN encoders, demonstrating the power of combining varied data sources. Introduction of Visual Transformers: The thesis introduces innovative applications of visual transformers in traffic modeling. The use of the Contextual Vision Transformer network (C-ViT) is a significant advancement, enabling spatial-temporal forecasting of traffic accident risks with higher precision and accuracy, achieving state-of-the-art results and improving upon it through various modifications. This novel application of visual transformers represents a major step forward in traffic incident analysis. Segmentation and Analysis of Traffic Disruptions: Another major contribution is the development of new methods for segmenting traffic disruptions and associating these disruptions with specific accident reports. This includes an in-depth analysis of the impacts of such disruptions on traffic flow and speed, providing valuable insights into the dynamics of traffic incidents. Overall, the thesis presents a series of interconnected methodologies that collectively enhance our understanding of traffic incident dynamics. By offering a universal framework and innovative approaches, this research not only contributes to the field of traffic incident modeling and prediction but also opens new avenues for future research and development in traffic management strategies.","abstract_html":"The progress of global urbanization and growth of vehicular traffic have led to an increase in traffic incidents, increasing demand for efficient modeling and prediction methodologies essential for traffic management. To meet this challenge, the thesis proposes the application of advanced machine learning and deep learning techniques to enhance traffic incident modeling. The research contributions are outlined as follows: Development of a Universal Framework: A novel framework is introduced for predicting traffic incident duration across various road network layouts, handling classification and regression problems, addressing outliers and imbalanced data classes. This framework is designed to handle both classification and regression problems effectively, while also addressing the challenges of outliers and imbalanced data classes. It utilizes feature importance estimation methods, such as SHAP, to identify critical variables, and incorporates anomaly detection techniques like One-Class SVM and Isolation Forest into the prediction models. Exploration of Data Fusion Techniques: A key contribution of this research is the exploration of data fusion. By integrating different data types, including traffic flow information, textual incident descriptions, and historical traffic flow data, the thesis proposes a methodology to enhance incident duration prediction accuracy. This is achieved through the use of deep learning methodologies like LSTM and ANN encoders, demonstrating the power of combining varied data sources. Introduction of Visual Transformers: The thesis introduces innovative applications of visual transformers in traffic modeling. The use of the Contextual Vision Transformer network (C-ViT) is a significant advancement, enabling spatial-temporal forecasting of traffic accident risks with higher precision and accuracy, achieving state-of-the-art results and improving upon it through various modifications. This novel application of visual transformers represents a major step forward in traffic incident analysis. Segmentation and Analysis of Traffic Disruptions: Another major contribution is the development of new methods for segmenting traffic disruptions and associating these disruptions with specific accident reports. This includes an in-depth analysis of the impacts of such disruptions on traffic flow and speed, providing valuable insights into the dynamics of traffic incidents. Overall, the thesis presents a series of interconnected methodologies that collectively enhance our understanding of traffic incident dynamics. By offering a universal framework and innovative approaches, this research not only contributes to the field of traffic incident modeling and prediction but also opens new avenues for future research and development in traffic management strategies.","abstract_has_math":false,"creators":["Grigorev, Artur"],"institution":null,"degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2023,"date_issued":"2023","date_published":"2023","updated_at":"2026-07-24T06:32:27Z","subjects":[],"languages":["en_US"],"rights":["info:eu-repo/semantics/openAccess","The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.","© 2023 Artur Grigorev","au.edu.uts.lib/cph"],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/10453/177936","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Grigorev, Artur"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2024-04-16T01:21:38Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2024-04-16T01:21:38Z"]},{"key":"dc:date.issued","label":"Date","values":["2023"]},{"key":"dc:relation","label":"Dc Relation","values":["https://opus.lib.uts.edu.au/bitstream/10453/177936/1/thesis.pdf"]},{"key":"dc:type","label":"Dc Type","values":["Thesis"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en_US"]},{"key":"dc:rights","label":"Dc Rights","values":["info:eu-repo/semantics/openAccess","The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.","© 2023 Artur Grigorev","au.edu.uts.lib/cph"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["http://hdl.handle.net/10453/177936"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description","label":"Description","values":["University of Technology Sydney. Faculty of Engineering and Information Technology."]},{"key":"dc:description.abstract","label":"Abstract","values":["The progress of global urbanization and growth of vehicular traffic have led to an increase in traffic incidents, increasing demand for efficient modeling and prediction methodologies essential for traffic management. To meet this challenge, the thesis proposes the application of advanced machine learning and deep learning techniques to enhance traffic incident modeling. The research contributions are outlined as follows: Development of a Universal Framework: A novel framework is introduced for predicting traffic incident duration across various road network layouts, handling classification and regression problems, addressing outliers and imbalanced data classes. This framework is designed to handle both classification and regression problems effectively, while also addressing the challenges of outliers and imbalanced data classes. It utilizes feature importance estimation methods, such as SHAP, to identify critical variables, and incorporates anomaly detection techniques like One-Class SVM and Isolation Forest into the prediction models. Exploration of Data Fusion Techniques: A key contribution of this research is the exploration of data fusion. By integrating different data types, including traffic flow information, textual incident descriptions, and historical traffic flow data, the thesis proposes a methodology to enhance incident duration prediction accuracy. This is achieved through the use of deep learning methodologies like LSTM and ANN encoders, demonstrating the power of combining varied data sources. Introduction of Visual Transformers: The thesis introduces innovative applications of visual transformers in traffic modeling. The use of the Contextual Vision Transformer network (C-ViT) is a significant advancement, enabling spatial-temporal forecasting of traffic accident risks with higher precision and accuracy, achieving state-of-the-art results and improving upon it through various modifications. This novel application of visual transformers represents a major step forward in traffic incident analysis. Segmentation and Analysis of Traffic Disruptions: Another major contribution is the development of new methods for segmenting traffic disruptions and associating these disruptions with specific accident reports. This includes an in-depth analysis of the impacts of such disruptions on traffic flow and speed, providing valuable insights into the dynamics of traffic incidents. Overall, the thesis presents a series of interconnected methodologies that collectively enhance our understanding of traffic incident dynamics. By offering a universal framework and innovative approaches, this research not only contributes to the field of traffic incident modeling and prediction but also opens new avenues for future research and development in traffic management strategies."]},{"key":"dc:format","label":"Dc Format","values":["Thesis (PhD)"]},{"key":"dc:title","label":"Title","values":["Advanced AI techniques for comprehensive traffic incident analysis: enhancing incident duration prediction and accident risk forecasting"]}]}],"canonical_facts":{"dc:creator":["Grigorev, Artur"],"dc:date.accessioned":["2024-04-16T01:21:38Z"],"dc:date.available":["2024-04-16T01:21:38Z"],"dc:date.issued":["2023"],"dc:description":["University of Technology Sydney. Faculty of Engineering and Information Technology."],"dc:description.abstract":["The progress of global urbanization and growth of vehicular traffic have led to an increase in traffic incidents, increasing demand for efficient modeling and prediction methodologies essential for traffic management. To meet this challenge, the thesis proposes the application of advanced machine learning and deep learning techniques to enhance traffic incident modeling. The research contributions are outlined as follows: Development of a Universal Framework: A novel framework is introduced for predicting traffic incident duration across various road network layouts, handling classification and regression problems, addressing outliers and imbalanced data classes. This framework is designed to handle both classification and regression problems effectively, while also addressing the challenges of outliers and imbalanced data classes. It utilizes feature importance estimation methods, such as SHAP, to identify critical variables, and incorporates anomaly detection techniques like One-Class SVM and Isolation Forest into the prediction models. Exploration of Data Fusion Techniques: A key contribution of this research is the exploration of data fusion. By integrating different data types, including traffic flow information, textual incident descriptions, and historical traffic flow data, the thesis proposes a methodology to enhance incident duration prediction accuracy. This is achieved through the use of deep learning methodologies like LSTM and ANN encoders, demonstrating the power of combining varied data sources. Introduction of Visual Transformers: The thesis introduces innovative applications of visual transformers in traffic modeling. The use of the Contextual Vision Transformer network (C-ViT) is a significant advancement, enabling spatial-temporal forecasting of traffic accident risks with higher precision and accuracy, achieving state-of-the-art results and improving upon it through various modifications. This novel application of visual transformers represents a major step forward in traffic incident analysis. Segmentation and Analysis of Traffic Disruptions: Another major contribution is the development of new methods for segmenting traffic disruptions and associating these disruptions with specific accident reports. This includes an in-depth analysis of the impacts of such disruptions on traffic flow and speed, providing valuable insights into the dynamics of traffic incidents. Overall, the thesis presents a series of interconnected methodologies that collectively enhance our understanding of traffic incident dynamics. By offering a universal framework and innovative approaches, this research not only contributes to the field of traffic incident modeling and prediction but also opens new avenues for future research and development in traffic management strategies."],"dc:format":["Thesis (PhD)"],"dc:identifier.uri":["http://hdl.handle.net/10453/177936"],"dc:language.iso":["en_US"],"dc:relation":["https://opus.lib.uts.edu.au/bitstream/10453/177936/1/thesis.pdf"],"dc:rights":["info:eu-repo/semantics/openAccess","The author owns the copyright in this thesis including all reproduction and reuse rights for the work. The work may not be altered without the permission of the copyright owner. Attribution is essential when quoting or paraphrasing from this thesis.","© 2023 Artur Grigorev","au.edu.uts.lib/cph"],"dc:title":["Advanced AI techniques for comprehensive traffic incident analysis: enhancing incident duration prediction and accident risk forecasting"],"dc:type":["Thesis"]},"updated_at":"2026-07-24T06:32:27Z"}