{"id":{"repo_id":"ecu","oai_identifier":"oai:thescholarship.ecu.edu:10342/14041"},"canonical_url":"https://search.dev.ndltd.org/etd/ecu/oai:thescholarship.ecu.edu:10342/14041","repository":{"repo_id":"ecu","name":"East Carolina University","base_url":"https://thescholarship.ecu.edu/server/oai/request"},"display":{"title":"IMPROVING MULTI-VARIATE TIME SERIES FORECASTING WITH DYNAMIC MULTI-HEAD ATTENTION ADJACENCY MATRIX","abstract":"Time series data is prevalent in many fields, such as finance, weather forecasting, and economics. Predicting future values of a time series can offer valuable insights for decision-making, such as identifying trends, detecting anomalies, and improving resource allocation. Existing research, including neural network-based models and transformer-based models, has demonstrated high performance in learning temporal information. However, capturing spatial information within time series data remains a significant challenge. In this project, we explored whether the attention mechanism can be effectively integrated into non-transformer-based models to enhance their ability to learn spatial information. To achieve this goal, we propose a novel framework that uses a dynamically learned adjacency matrix based on related work called the Multi-variate Time-series Graph Neural Network(MTGNN). Instead of using a correlation-based learned adjacency matrix, the adjacency matrix and graph modules are replaced with a dynamically learned adjacency matrix with multi-attention. This framework shows that a dynamically learned attention adjacency matrix can perform as well as other frameworks when learning spatial information.","abstract_html":"Time series data is prevalent in many fields, such as finance, weather forecasting, and economics. Predicting future values of a time series can offer valuable insights for decision-making, such as identifying trends, detecting anomalies, and improving resource allocation. Existing research, including neural network-based models and transformer-based models, has demonstrated high performance in learning temporal information. However, capturing spatial information within time series data remains a significant challenge. In this project, we explored whether the attention mechanism can be effectively integrated into non-transformer-based models to enhance their ability to learn spatial information. To achieve this goal, we propose a novel framework that uses a dynamically learned adjacency matrix based on related work called the Multi-variate Time-series Graph Neural Network(MTGNN). Instead of using a correlation-based learned adjacency matrix, the adjacency matrix and graph modules are replaced with a dynamically learned adjacency matrix with multi-attention. This framework shows that a dynamically learned attention adjacency matrix can perform as well as other frameworks when learning spatial information.","abstract_has_math":false,"creators":["Bruce, Ashley Denise"],"institution":"East Carolina University","degree_name":null,"degree_level":null,"degree_discipline":null,"degree_department":"Computer Science","school":null,"contributors":[],"advisors":["Ding, Qin"],"committee_chairs":[],"committee_members":["Dr. Sinan Sousan","Dr. Nic Herndon"],"year":2025,"date_issued":"May 2025","date_published":"May 2025","updated_at":"2026-07-24T02:13:44Z","subjects":["attention"],"languages":[],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://hdl.handle.net/10342/14041","outbound_label":"Handle","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:contributor.advisor","label":"Advisor","values":["Ding, Qin"]},{"key":"dc:contributor.committeemember","label":"Committee Member","values":["Dr. Sinan Sousan","Dr. Nic Herndon"]},{"key":"dc:contributor.department","label":"Department","values":["Computer Science"]},{"key":"dc:contributor.other","label":"Dc Contributor Other","values":["Computer Science"]},{"key":"dc:creator","label":"Author","values":["Bruce, Ashley Denise"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2025-06-05T17:30:22Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2025-06-05T17:30:22Z"]},{"key":"dc:date.issued","label":"Date","values":["May 2025"]},{"key":"dc:publisher","label":"Institution","values":["East Carolina University"]},{"key":"dc:type","label":"Dc Type","values":["Master's Thesis"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["attention"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["http://hdl.handle.net/10342/14041"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["Time series data is prevalent in many fields, such as finance, weather forecasting, and economics. Predicting future values of a time series can offer valuable insights for decision-making, such as identifying trends, detecting anomalies, and improving resource allocation. Existing research, including neural network-based models and transformer-based models, has demonstrated high performance in learning temporal information. However, capturing spatial information within time series data remains a significant challenge. In this project, we explored whether the attention mechanism can be effectively integrated into non-transformer-based models to enhance their ability to learn spatial information. To achieve this goal, we propose a novel framework that uses a dynamically learned adjacency matrix based on related work called the Multi-variate Time-series Graph Neural Network(MTGNN). Instead of using a correlation-based learned adjacency matrix, the adjacency matrix and graph modules are replaced with a dynamically learned adjacency matrix with multi-attention. This framework shows that a dynamically learned attention adjacency matrix can perform as well as other frameworks when learning spatial information."]},{"key":"dc:format.mimetype","label":"Dc Format Mimetype","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["IMPROVING MULTI-VARIATE TIME SERIES FORECASTING WITH DYNAMIC MULTI-HEAD ATTENTION ADJACENCY MATRIX"]}]}],"canonical_facts":{"dc:contributor.advisor":["Ding, Qin"],"dc:contributor.committeemember":["Dr. Sinan Sousan","Dr. Nic Herndon"],"dc:contributor.department":["Computer Science"],"dc:contributor.other":["Computer Science"],"dc:creator":["Bruce, Ashley Denise"],"dc:date.accessioned":["2025-06-05T17:30:22Z"],"dc:date.available":["2025-06-05T17:30:22Z"],"dc:date.issued":["May 2025"],"dc:description.abstract":["Time series data is prevalent in many fields, such as finance, weather forecasting, and economics. Predicting future values of a time series can offer valuable insights for decision-making, such as identifying trends, detecting anomalies, and improving resource allocation. Existing research, including neural network-based models and transformer-based models, has demonstrated high performance in learning temporal information. However, capturing spatial information within time series data remains a significant challenge. In this project, we explored whether the attention mechanism can be effectively integrated into non-transformer-based models to enhance their ability to learn spatial information. To achieve this goal, we propose a novel framework that uses a dynamically learned adjacency matrix based on related work called the Multi-variate Time-series Graph Neural Network(MTGNN). Instead of using a correlation-based learned adjacency matrix, the adjacency matrix and graph modules are replaced with a dynamically learned adjacency matrix with multi-attention. This framework shows that a dynamically learned attention adjacency matrix can perform as well as other frameworks when learning spatial information."],"dc:format.mimetype":["application/pdf"],"dc:identifier.uri":["http://hdl.handle.net/10342/14041"],"dc:publisher":["East Carolina University"],"dc:subject":["attention"],"dc:title":["IMPROVING MULTI-VARIATE TIME SERIES FORECASTING WITH DYNAMIC MULTI-HEAD ATTENTION ADJACENCY MATRIX"],"dc:type":["Master's Thesis"]},"updated_at":"2026-07-24T02:13:44Z"}