Paper: arXiv 2508.06497
Authors: Mohammed-Khalil Ghali, Cecil Pang, Oscar Molina, Carlos Gershenson-Garcia, Daehan Won
Abstract
Accurate forecasting of commodity price spikes is vital for countries with limited economic buffers, where sudden increases can strain national budgets, disrupt import-reliant sectors, and undermine food and energy security. This paper introduces a hybrid forecasting framework that combines historical commodity price data with semantic signals derived from global economic news, using an agentic generative AI pipeline. The architecture integrates dual-stream Long Short-Term Memory (LSTM) networks with attention mechanisms to fuse structured time-series inputs with semantically embedded, fact-checked news summaries collected from 1960 to 2023. The model is evaluated on a 64-year dataset comprising normalized commodity price series and temporally aligned news embeddings. Results show that the proposed approach achieves a mean AUC of 0.94 and an overall accuracy of 0.91 substantially outperforming traditional baselines such as logistic regression (AUC = 0.34), random forest (AUC = 0.57), and support vector machines (AUC = 0.47). Additional ablation studies reveal that the removal of attention or dimensionality reduction leads to moderate declines in performance, while eliminating the news component causes a steep drop in AUC to 0.46, underscoring the critical value of incorporating real-world context through unstructured text. These findings demonstrate that integrating agentic generative AI with deep learning can meaningfully improve early detection of commodity price shocks, offering a practical tool for economic planning and risk mitigation in volatile market environments while saving the very high costs of operating a full generative AI agents pipeline.
Complexity vs Empirical Score
- Math Complexity: 6.5/10
- Empirical Rigor: 8.5/10
- Quadrant: Holy Grail — high math complexity, high empirical rigor
Why this score: The paper employs advanced deep learning architectures (dual-stream LSTM with attention, news embeddings) and statistical evaluation (AUC, ablation studies), while demonstrating substantial empirical rigor through a 64-year dataset, real-world data integration, and direct comparisons against established baselines.
Research Flowchart
flowchart TD A["Research Goal<br>Forecast Commodity Price Shocks"] --> B["Data & Inputs<br>Prices (1960-2023) & Economic News"] B --> C["Methodology<br>Agentic GenAI for News Extraction"] C --> D["Computational Process<br>Dual-Stream LSTM + Attention Fusion"] D --> E["Key Outcomes<br>High Accuracy (0.91) & AUC (0.94)"] E --> F["Insight<br>News Context is Critical for Performance"]
Evidence Card
What the authors claim, on what data, and who has checked it. Read from the paper's own text by a language model: every quote is verbatim, nothing was re-run. How to read this card
Proposes
The paper proposes a novel hybrid forecasting framework that fuses historical commodity price data with semantic news signals using an agentic generative AI pipeline, implemented as a dual‑stream LSTM with attention mechanisms.
Source sentence
This paper introduces a hybrid forecasting framework that combines historical commodity price data with semantic signals derived from global economic news, using an agentic generative AI pipeline.
Assumes 3
- The analysis is performed on yearly aggregated data, limiting responsiveness to short‑term events. data
Source sentence
The analysis is based on yearly data, which limits the model’s responsiveness to short-term events.
- News summaries produced by the generative AI agent are fact‑checked and accurate. model
Source sentence
semantically embedded, fact‑checked news summaries collected from 1960 to 2023.
- A spike year is defined when the year‑over‑year percentage change exceeds a predefined threshold. statistical
Source sentence
A year yi is labeled as a spike year if the relative change in average price from the previous year exceeds a predefined thr…
Data
- World Bank historical commodity price data market-prices public
Source sentence
This study utilizes publicly available historical commodity price data provided by the World Bank [31], covering the period from 1960 to 2023.
- Agentic generative AI news embeddings text-news unknown
Source sentence
economic news summaries generated through the agentic generative AI pipeline are converted into dense vector representations using sentence‑level embedding models.
Costs
Not applicable
Finds 3
- The proposed model achieves a mean AUC of 0.94 and overall accuracy of 0.91 on 64 years of data, substantially outperforming traditional baselines. predictive-accuracy
Source sentence
Results show that the proposed approach achieves a mean AUC of 0.94 and an overall accuracy of 0.91 substantially outperforming traditional baselines such as logistic regression (AUC = 0.34), random forest (AUC = 0.57), and support vector machines (AUC = 0.47).
- Ablation removing the news component causes AUC to drop sharply to 0.46, highlighting the critical value of semantic news context. performance
Source sentence
eliminating the news component causes a steep drop in AUC to 0.46, underscoring the critical value of incorporating real‑world context through unstructured text.
- Removing attention reduces mean AUC to 0.8263, while removing PCA reduces it to 0.7584, showing each component contributes to performance. performance
Source sentence
Without Attention 0.8263 ± 0.0889; Without PCA 0.7584 ± 0.1302.
Checked by others
No other paper in the archive has been found to support or contradict these findings (compared with 17 related papers). Coverage is partial, so this is neither replication nor refutation.
Code
No repository by the authors found in the paper.
Extracted 2026-11-14 · same card as JSON · A language model chose these items: check the source sentence before relying on one.
Cited References (33)
- Crude Oil Price Forecasting Techniques: A Comprehensive Review of Literature (2013)
- Forecasting the Real Price of Oil in a Changing World: A Forecast Combination Approach (2015)
- An Assessment of Time Series Methods in Metal Price Forecasting (2005)
- Forecasting the COMEX Copper Spot Price by Means of Neural Networks and ARIMA Models (2015)
- Forecasting Crude Oil Prices: A Comparison Between Artificial Neural Networks and Vector Autoregressive Models (2019)
- Forecasting Commodity Prices: Looking for a Benchmark (2021)
- Financial Time Series Forecasting with Deep Learning : A Systematic Literature Review: 2005–2019 (2020)
- A Deep Learning Ensemble Approach for Crude Oil Price Forecasting (2017)
and 25 more
- Openai/openai-agents-python: A Lightweight, powerful framework for multi-agent workflows — link
- Evolutionary Neural Network Model for West Texas Intermediate Crude Oil Price Prediction (2015)
- Forecasting Copper Price by Application of Robust Artificial Intelligence Techniques (2021)
- Impact of Media Hype and Fake News on Commodity Futures Prices: A Deep Learning Approach over the COVID-19 Period (2024)
- Giving Content to Investor Sentiment: The Role of Media in the Stock Market (2007)
- Text Mining for Market Prediction: A Systematic Review (2014)
- Deep Learning for Event-Driven Stock Prediction (2015)
- Commodity Markets (2024) — link
- The Role of News Sentiment in Oil Futures Returns and Volatility Forecasting: A Data-Decomposition Based Deep Learning Approach (2021)
- A Deep Fusion Model for Stock Market Prediction with News Headlines and Time Series Data (2024)
- FinBERT: A Pre-trained Financial Language Representation Model for Financial Text Mining (2020)
- Can ChatGPT Forecast Stock Price Movements? Return Predictability and Large Language Models — → 2304.07619
- BloombergGPT: A Large Language Model for Finance (2023) — link
- FinAgent: A Multimodal Foundation Agent for Financial Trading — → 2402.18485
- TradingGPT: Multi-Agent System with Layered Memory and Distinct Characters for Enhanced Financial Trading Performance — → 2309.03736
- LLMFactor: Extracting Profitable Factors through Prompts for Explainable Stock Movement Prediction (2024) — link
- Multi-modal Time Series Analysis: A Tutorial and Survey (2024) — link
- BEYONDWORDS is All You Need: Agentic Generative AI based Social Media Themes Extractor (2025) — link
- Adding fuel to the fire: cheap oil during the COVID-19 pandemic (2020)
- Agricultural growth and poverty reduction: Additional evidence (2010)
- What’s driving food prices in 2011? (2011)
- Role of international price and domestic inflation in triggering export restrictions on food commodities (2024)
- Breaks, trends and unit roots in commodity prices: a robust investigation (2014)
- Exploring the Trend of Commodity Prices: A review and bibliometric analysis (2022)
- Trade balance, monetary supply, commodity prices, and greener energy growth: Contextual evidence from BRICS economies in the lens of sustainability (2023)
3 of 33 cited works are in the corpus; resolved via arXiv source bundles, Semantic Scholar and OpenAlex. Corpus-wide: 2596 linked papers.