AI Forecasting

In recent years, the rapid development of artificial intelligence (AI) technologies, especially cutting-edge methods such as deep learning, generative large language models, and multimodal foundational models, has profoundly reshaped the theoretical framework and methodology of forecasting science. These technologies have not only significantly enhanced the accuracy and efficiency of forecasting tasks but also greatly expanded the scope and dimensions of forecasting targets, covering everything from traditional structured time series to unstructured text, images, and even cross-modal integrated data. AI-based forecasting methods have demonstrated unprecedented potential in key areas such as complex system modeling, uncertainty quantification, explainable forecasting, and real-time dynamic forecasting, gradually becoming a core infrastructure supporting scientific decision-making and intelligent management.

From macroeconomic and financial market volatility forecasting to long-term simulations of climate and environmental changes, and even early disease warning in the healthcare sector, AI-driven forecasting research is driving paradigm shifts across multiple critical fields. Particularly, with the emergence of new methods such as dedicated forecasting models, neural differential equations, spatiotemporal attention mechanisms, and generative-enhanced inference, forecasting science has entered a new developmental stage that integrates data and knowledge-driven high-dimensional dynamic modeling with explainable AI.

In this context, the "AI forecasting" column aims to promote original innovations, interdisciplinary method integration, and practical application verification in this field. We specifically focus on frontier topics such as the construction of next-generation AI models for complex forecasting scenarios, the design of scalable and trustworthy forecasting algorithms, uncertainty quantification in forecasting systems, and human–machine collaborative decision-making mechanisms. We also encourage forward-thinking and practical research outcomes in key areas such as economic and social governance, sustainable environmental development, and smart cities.

This column aims to build a high-level international academic exchange platform dedicated to advancing the innovative development and deep integration of cutting-edge artificial intelligence technologies in forecasting science and applications. We encourage submissions focused on areas such as machine learning, deep learning, generative models, spatiotemporal forecasting models, multimodal learning, uncertainty modeling, and explainable AI. Submissions should cover both theoretical innovations and cross-disciplinary practical applications, including but not limited to fields like finance and economics, meteorology and climate, energy dispatch, healthcare, transportation logistics, and supply chain management.

This column aligns closely with the overall scope of Forecasting, both being committed to disseminating the latest research progress in forecasting science. However, our focus is more specifically on how artificial intelligence, as a disruptive technology, is expanding and reshaping the modern forecasting methodology system. We aim to foster deep interaction and collaboration between the forecasting science and AI communities, providing scholars with a dedicated platform for releasing cutting-edge results and engaging in intellectual exchange, collectively driving the future development of predictive intelligence.

This column invites authors to submit their work to promote theoretical innovation and interdisciplinary applications of artificial intelligence (AI) technologies in the field of forecasting, fostering the deep integration and collaborative development of forecasting science and AI.

Potential topics include, but are not limited to, the following areas:

• Fundamental theories and new methodological frameworks for AI-driven forecasting;
• Systematic applications of generative large language models in forecasting;
• Integration of multimodal foundational models with forecasting;
• Pretrained foundational models for forecasting tasks;
• Long-sequence time series forecasting algorithms;
• Dynamic inference and adaptive forecasting systems;
• Cutting-edge applications of few-shot/zero-shot learning in forecasting tasks;
• Unified processing and feature extraction of heterogeneous forecasting data;
• Computer vision-driven forecasting tasks;
• Natural language processing and forecasting integration;
• Applications of reinforcement learning in dynamic forecasting and sequential decision-making;
• Explainability, uncertainty quantification, and trustworthiness assessment of AI forecasting models;
• Efficiency optimization and system deployment of forecasting models;
• Development of evaluation metrics and benchmark testing frameworks for AI forecasting;
• Ethical issues, algorithmic fairness, and privacy protection in forecasting tasks;
• Hybrid forecasting paradigms combining data-driven and model-driven approaches;
• Frontier AI forecasting applications across domains.

Please note that submissions not explicitly focused on AI-driven forecasting or lacking theoretical research and practical applications targeted at forecasting tasks will not be considered for inclusion in this section.