[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-1-en-105":3,"doc-seo-564838-105":53,"doc-detail-564838-en":119},{"code":4,"msg":5,"data":6},0,"success",[7,14,19,24,29,34,39,44,49],{"id":8,"doc_module":9,"doc_module_name":10,"category_name":11,"show_sort_weight":12,"slug":13},11,1,"Template","Presentations",90,"presentations",{"id":15,"doc_module":9,"doc_module_name":10,"category_name":16,"show_sort_weight":17,"slug":18},12,"Resumes",80,"resumes",{"id":20,"doc_module":9,"doc_module_name":10,"category_name":21,"show_sort_weight":22,"slug":23},14,"Invoices",70,"invoices",{"id":25,"doc_module":9,"doc_module_name":10,"category_name":26,"show_sort_weight":27,"slug":28},15,"Posters",60,"posters",{"id":30,"doc_module":9,"doc_module_name":10,"category_name":31,"show_sort_weight":32,"slug":33},16,"Social Media",50,"social-media",{"id":35,"doc_module":9,"doc_module_name":10,"category_name":36,"show_sort_weight":37,"slug":38},17,"Forms",40,"forms",{"id":40,"doc_module":9,"doc_module_name":10,"category_name":41,"show_sort_weight":42,"slug":43},18,"Letters",30,"letters",{"id":45,"doc_module":9,"doc_module_name":10,"category_name":46,"show_sort_weight":47,"slug":48},21,"Paper Templates",5,"papers-templates",{"id":50,"doc_module":9,"doc_module_name":10,"category_name":51,"show_sort_weight":4,"slug":52},158,"General","general-158",{"code":4,"msg":54,"data":55},"ok",{"site_id":56,"language":57,"slug":58,"title":59,"keywords":60,"description":61,"schema_data":62,"social_meta":112,"head_meta":114,"extra_data":116,"updated_unix":118},105,"en","wti-futures-price-forecasting-based-on-multi-graph-fusion-spatiotemporal-attention-network","WTI Futures Price Forecasting Based on Multi-Graph Fusion Spatiotemporal Attention Network","","This study develops a Multi-Graph Fusion Spatiotemporal Attention Network (MG-STAN) to model evolving interactions between crude oil markets and related financial systems for futures price forecasting. The framework integrates temporal embeddings, spatial attention modules, and a multi-graph structure to represent heterogeneous inter-market relationships. Using data spanning 2011–2024 with commodity futures, supply-demand factors, and financial indicators, MG-STAN outperforms conventional deep learning baselines. A three-graph fusion strategy combining correlation, K-nearest neighbor, and dynamic time warping graphs yields the best results.",{"@graph":63,"@context":111},[64,80,102],{"@type":65,"itemListElement":66},"BreadcrumbList",[67,71,74,77],{"item":68,"name":69,"@type":70,"position":9},"https://docshare.wps.com","Home","ListItem",{"item":72,"name":10,"@type":70,"position":73},"https://docshare.wps.com/template/",2,{"item":75,"name":51,"@type":70,"position":76},"https://docshare.wps.com/template/general/",3,{"item":78,"name":59,"@type":70,"position":79},"https://docshare.wps.com/template/wti-futures-price-forecasting-based-on-multi-graph-fusion-spatiotemporal-attention-network/564838/",4,{"url":78,"name":59,"@type":81,"image":82,"author":87,"headline":59,"publisher":90,"fileFormat":93,"inLanguage":57,"description":61,"dateModified":94,"datePublished":95,"encodingFormat":93,"isAccessibleForFree":96,"interactionStatistic":97},"DigitalDocument",{"url":83,"@type":84,"width":85,"height":86},"https://docshare.wps.com/thumbnails/wti-futures-price-forecasting-based-on-multi-graph-fusion-spatiotemporal-attention-network/564838.png","ImageObject",442,249,{"name":88,"@type":89},"Ben ","Person",{"url":68,"name":91,"@type":92},"DocShare","Organization","application/pdf","2026-10-22","2026-10-12",true,{"@type":98,"interactionType":99,"userInteractionCount":101},"InteractionCounter",{"@type":100},"ViewAction",7,{"@type":103,"mainEntity":104},"FAQPage",[105],{"name":106,"@type":107,"acceptedAnswer":108},"Which graph fusion strategy performs best in the study?","Question",{"text":109,"@type":110},"The three-graph fusion strategy integrating correlation, K-nearest neighbor, and dynamic time warping graphs achieves the best forecasting results.","Answer","https://schema.org",{"og:url":78,"og:type":113,"og:title":59,"og:site_name":91,"og:description":61},"article",{"robots":115,"canonical":78},"index,follow",{"doc_id":117,"site_id":56},564838,1792114189,{"code":4,"msg":5,"data":120},{"doc_id":117,"user_id":121,"nickname":88,"user_avatar":122,"doc_module":9,"category_id":50,"category_name":51,"doc_title":59,"doc_description":61,"doc_content":123,"file_id":124,"file_url":125,"file_type":126,"file_size":127,"view_count":101,"is_deleted":4,"is_public":9,"is_downloadable":9,"audit_status":9,"page_count":128,"language":129,"language_code":57,"site_id":56,"html_lang":57,"table_of_contents":130,"faqs":131,"seo_title":132,"seo_description":61,"update_tm":133,"read_time":79},2336478951081,"https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c","WTI Futures Price Forecasting Based on Multi-Graph Fusion Spatiotemporal Attention Network  \nJunke Huanga*, Hui Qub  \nSchool of Management and Engineering, Nanjing University, Nanjing, China  \na* [1287171434@qq.com](1287171434@qq.com), [b](b linda59qu@nju.edu.cn)[ linda59qu@nju.edu.cn](b linda59qu@nju.edu.cn)  \nAbstract. This study develops a Multi-Graph Fusion Spatiotemporal Attention Network (MG-STAN) to better capture the evolving interactions between crude oil markets and related financial systems. The proposed framework incorporates temporal embeddings, spatial attention modules, and a multi-graph structure to reflect diverse inter-market relationships. Using a dataset covering 2011–2024 that includes commodity futures, supply-demand factors, and financial indicators, our proposed MG-STAN models consistently and significantly outperform conventional deep learning models. Notably, a three-graph fusion strategy—combining correlation, K-nearest neighbor and dynamic time warping graphs—achieves the best results, suggesting that selectively integrating heterogeneous graphs can enhance forecasting accuracy. The findings underscore the value of multi-graph designs and attention mechanisms in modeling market complexity, and offer new perspectives for price forecasting and energy finance research.  \nKeywords: Crude oil futures, Price forecast, Spatiotemporal graph neural network, Multi-graph fusion  \n1 Introduction  \nCrude oil is pivotal to global energy systems, with price fluctuations significantly affecting economic stability. The futures market serves as a key venue for price discovery and risk management[1]. This study focuses on enhancing crude oil futures price forecasting precision through advanced spatiotemporal modeling.  \nForecasting methodologies have evolved from linear econometric models (e.g., ARIMA, GARCH)[2,3], limited by non-stationarity and nonlinearity, to deep learning approaches. ANNs, RNN variants (LSTM, GRU), and CNNs capture temporal patterns and multi-scale features[4,5]. GNNs model inter-market relationships via graph representations citep{GNN1}, with spatiotemporal variants (e.g., STGCNs) jointly encoding spatial and temporal dependencies[6]. However, financial markets lack intrinsic graph structures. While static graphs rely on predefined relationships, adaptive methods (e.g., GAT) learn dynamic connections[7,8] . Multi-graph frameworks integrate diverse topologies to better represent complex spatial dependencies, showing promise in financial forecasting[9,10].  \n© The Author(s) 2026  \nD. Magni et al. (eds.), Proceedings of the 2026 3rd International Conference on Applied Economics, Management Science and Social Development (AEMSS 2026), Advances in Economics, Business and Management Research 389,  \n[https://doi.org/10.2991/978-94-6239-672-2_35](https://doi.org/10.2991/978-94-6239-672-2_35)  \nTo address these challenges, we propose MG-STAN (Multi-Graph Fusion Gated Recurrent Spatial Attention Network), comprising: (1) a temporal embedding module for short- and long-term dependencies; (2) a spatial attention module using GAT and Transformers; (3) a multi-graph fusion module integrating heterogeneous graph structures; and (4) a convolution-based prediction module. This work advances GNN-based spatiotemporal modeling for crude oil markets.  \nMain contributions:  \n(1) Proposes MG-STAN integrating attention mechanisms and multi-graph fusion for precise crude oil price forecasting.  \n(2) Constructs multiple graph structures to capture diverse market relationships from complementary perspectives.  \n(3) Designs an effective fusion layer that integrates multi-graph information, significantly improving prediction accuracy over single-graph baselines.  \nThe paper is organized as follows: Section 2 details the MG-STAN framework; Section 3 describes the dataset and experimental setup; Section 4 presents results and analysis; Section 5 concludes with future directions.  \n2 Proposed Method  \nIn Section 2.1, we will f","cbCaiqsDGkOUZTGG","https://ap.wps.com/l/cbCaiqsDGkOUZTGG","pdf",1586931,10,"English","# Introduction\n## Forecasting background and related methods\n# Proposed Method\n## Problem description\n## Graph construction\n### Correlation coefficient matrix\n### Granger causality matrix\n### K-nearest neighbor matrix","[{\"question\":\"Which graph fusion strategy performs best in the study?\",\"answer\":\"The three-graph fusion strategy integrating correlation, K-nearest neighbor, and dynamic time warping graphs achieves the best forecasting results.\"}]","WTI Futures Price Forecasting Based on Multi-Graph Fusion Spatiotemporal Attention Network | PDF",1791844503]