Traffic congestion is not foreign to major metropolitan areas. Congestion in large cities often is associated with dense land developments and continued economic growth. In general, congestion can be classified into two categories: recurring and nonrecurring. Recurring congestion often occurs at certain parts of highway networks, referred to as bottleneck locations. Nonrecurring congestion, on the other hand, can be caused by different reasons, including work zones, special events, accidents, inclement weather, poor signal timing, etc. The work presented here demonstrates an approach to effectively identifying spatiotemporal patterns of traffic congestion at a network level. The Metro Atlanta highway network was used as a case study. Real time traffic data was acquired from the Georgia Department of Transportation (GDOT) Navigator system. For a qualitative analysis, speed data was categorized into three levels: low, median, and high. Cluster analysis was performed with respect to the categorized speed data in the spatiotemporal domain to identify where and when congestion has occurred and for how long, which indicate the severity of congestion. This qualitative analysis was performed by day of week to identify potential variation in congestion over weekdays and weekend. For a quantitative analysis, actual speed data was used to construct daily spatiotemporal maps to reveal congestion patterns at a more granular level, where congestion is represented as “cloud” in the spatiotemporal domain. Future work will be focusing on deep learning of congestion patterns using Convolutional Long Short Term Memory (ConvLSTM) networks.
","abstract_html":"<p>Traffic congestion is not foreign to major metropolitan areas. Congestion in large cities often is associated with dense land developments and continued economic growth. In general, congestion can be classified into two categories: recurring and nonrecurring. Recurring congestion often occurs at certain parts of highway networks, referred to as bottleneck locations. Nonrecurring congestion, on the other hand, can be caused by different reasons, including work zones, special events, accidents, inclement weather, poor signal timing, etc. The work presented here demonstrates an approach to effectively identifying spatiotemporal patterns of traffic congestion at a network level. The Metro Atlanta highway network was used as a case study. Real time traffic data was acquired from the Georgia Department of Transportation (GDOT) Navigator system. For a qualitative analysis, speed data was categorized into three levels: low, median, and high. Cluster analysis was performed with respect to the categorized speed data in the spatiotemporal domain to identify where and when congestion has occurred and for how long, which indicate the severity of congestion. This qualitative analysis was performed by day of week to identify potential variation in congestion over weekdays and weekend. For a quantitative analysis, actual speed data was used to construct daily spatiotemporal maps to reveal congestion patterns at a more granular level, where congestion is represented as “cloud” in the spatiotemporal domain. Future work will be focusing on deep learning of congestion patterns using Convolutional Long Short Term Memory (ConvLSTM) networks.</p>","abstract_has_math":false,"creators":["Kretlow, Betty"],"institution":null,"degree_name":"Master of Science in Computer Science (MSCS)","degree_level":"Thesis","degree_discipline":"Computer Science","degree_department":null,"school":null,"contributors":["Dr. Chih-Cheng Hung","Dr. Jidong Yang"],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2019,"date_issued":"2019-10-10T07:00:00Z","date_published":"2019-10-10T07:00:00Z","updated_at":"2026-07-24T02:43:33Z","subjects":["Traffic congestion; convolutional neural network; spatio-temporal analysis","Artificial Intelligence and Robotics"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"https://digitalcommons.kennesaw.edu/cs_etd/27","outbound_label":"Repository record","outbound_source":"dc:identifier"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor","label":"Contributor","values":["Dr. Chih-Cheng Hung","Dr. Jidong Yang"]},{"key":"dc:creator","label":"Author","values":["Kretlow, Betty"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.available","label":"Dc Date Available","values":["2019-10-10T07:00:00Z"]},{"key":"thesis:degree_discipline","label":"Discipline","values":["Computer Science"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Thesis"]},{"key":"thesis:degree_name","label":"Degree Name","values":["Master of Science in Computer Science (MSCS)"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Traffic congestion; convolutional neural network; spatio-temporal analysis","Artificial Intelligence and Robotics"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier","label":"Identifier","values":["https://digitalcommons.kennesaw.edu/cs_etd/27"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Traffic congestion is not foreign to major metropolitan areas. Congestion in large cities often is associated with dense land developments and continued economic growth. In general, congestion can be classified into two categories: recurring and nonrecurring. Recurring congestion often occurs at certain parts of highway networks, referred to as bottleneck locations. Nonrecurring congestion, on the other hand, can be caused by different reasons, including work zones, special events, accidents, inclement weather, poor signal timing, etc. The work presented here demonstrates an approach to effectively identifying spatiotemporal patterns of traffic congestion at a network level. The Metro Atlanta highway network was used as a case study. Real time traffic data was acquired from the Georgia Department of Transportation (GDOT) Navigator system. For a qualitative analysis, speed data was categorized into three levels: low, median, and high. Cluster analysis was performed with respect to the categorized speed data in the spatiotemporal domain to identify where and when congestion has occurred and for how long, which indicate the severity of congestion. This qualitative analysis was performed by day of week to identify potential variation in congestion over weekdays and weekend. For a quantitative analysis, actual speed data was used to construct daily spatiotemporal maps to reveal congestion patterns at a more granular level, where congestion is represented as “cloud” in the spatiotemporal domain. Future work will be focusing on deep learning of congestion patterns using Convolutional Long Short Term Memory (ConvLSTM) networks.
"]},{"key":"dc:title","label":"Title","values":["DEVELOPMENT OF SPATIOTEMPORAL CONGESTION PATTERN OBSERVATION MODEL USING HISTORICAL AND NEAR REAL TIME DATA"]}]}],"canonical_facts":{"dc:contributor":["Dr. Chih-Cheng Hung","Dr. Jidong Yang"],"dc:creator":["Kretlow, Betty"],"dc:date.available":["2019-10-10T07:00:00Z"],"dc:description.abstract":["Traffic congestion is not foreign to major metropolitan areas. Congestion in large cities often is associated with dense land developments and continued economic growth. In general, congestion can be classified into two categories: recurring and nonrecurring. Recurring congestion often occurs at certain parts of highway networks, referred to as bottleneck locations. Nonrecurring congestion, on the other hand, can be caused by different reasons, including work zones, special events, accidents, inclement weather, poor signal timing, etc. The work presented here demonstrates an approach to effectively identifying spatiotemporal patterns of traffic congestion at a network level. The Metro Atlanta highway network was used as a case study. Real time traffic data was acquired from the Georgia Department of Transportation (GDOT) Navigator system. For a qualitative analysis, speed data was categorized into three levels: low, median, and high. Cluster analysis was performed with respect to the categorized speed data in the spatiotemporal domain to identify where and when congestion has occurred and for how long, which indicate the severity of congestion. This qualitative analysis was performed by day of week to identify potential variation in congestion over weekdays and weekend. For a quantitative analysis, actual speed data was used to construct daily spatiotemporal maps to reveal congestion patterns at a more granular level, where congestion is represented as “cloud” in the spatiotemporal domain. Future work will be focusing on deep learning of congestion patterns using Convolutional Long Short Term Memory (ConvLSTM) networks.
"],"dc:identifier":["https://digitalcommons.kennesaw.edu/cs_etd/27"],"dc:subject":["Traffic congestion; convolutional neural network; spatio-temporal analysis","Artificial Intelligence and Robotics"],"dc:title":["DEVELOPMENT OF SPATIOTEMPORAL CONGESTION PATTERN OBSERVATION MODEL USING HISTORICAL AND NEAR REAL TIME DATA"],"thesis:degree_discipline":["Computer Science"],"thesis:degree_level":["Thesis"],"thesis:degree_name":["Master of Science in Computer Science (MSCS)"]},"updated_at":"2026-07-24T02:43:33Z"}