November 2025 – AI/ML in Commodities

This paper investigates the ability of a ChatGPT-based indicator to forecast excess returns of a commodity futures index. Using ChatGPT to extract information from over 2.5 million articles from nine international newspapers, the authors demonstrate that their constructed commodity news ratio index significantly predicts future commodity returns, both in-sample and out-of-sample. Furthermore, it outperforms traditional textual analysis methods while indicating economic significance within an asset allocation framework. The results highlight the critical role of ChatGPT in forecasting commodity market dynamics and provide valuable insights for both financial market participants and researchers.

The authors use prevailing machine learning models to investigate the predictability of futures returns in 22 commodities with commodity-specific and macroeconomic factors as predictors. Out-of-sample prediction errors for the majority of futures contracts are lowered compared with those obtained by the baseline models of AR(1) and forecast combinations. Using Shapley values to explain feature importance, we identify dominant predictors for each commodity. A long–short portfolio strategy based on monthly Light Gradient-boosting Machine predictions outperforms the benchmark linear models in terms of annual return, Sharpe ratio, and maximum drawdown.

This paper uses machine learning tools to study the lead-lag relations in commodity futures returns. The authors use LASSO (Least Absolute Shrinkage and Selection Operator) to select the predictors because the number of predictors is large relative to the number of observations. They find significant full-sample and out-of-sample predictability. The out-of-sample forecasts based on LASSO generate statistically and economically large gains. When the authors use more complex machine learning models such as regression trees and neural networks to forecast commodity futures returns, the out-of-sample performance is worse than LASSO portfolios, suggesting that nonlinearities and interactions do not appear substantial in the data. Finally, the authors find that index trading due to financialization has been associated with excess comovement among commodity futures.

This study introduces media emotion intensity—measuring the proportion of emotional versus factual content in news—as a novel predictor of commodity futures returns. Emotion intensity measures the proportion of emotional content relative to factual content in media news. The media emotion intensity factor generates an economically significant annual premium of approximately 13%, even after accounting for transaction costs. This premium is more pronounced for commodities with low media coverage, high momentum, high basis-momentum, high hedging pressure, and backwardation. Emotion intensity significantly predicts the trading tendencies of both commercial and non-commercial traders and the cross-section of commodity futures returns at both portfolio and individual levels. The study also finds that media emotion intensity predicts future commodities’ sentiment. Further, other commonly considered risk sources cannot subsume the predictability of the media emotion intensity factor.

Extending the popular heterogeneous autoregressive (HAR) model with additional information channels to forecast realized volatility of WTI futures prices, the authors show that machine learning-generated forecasts provide better forecasting quality and that portfolios that are constructed with these forecasts outperform their competing models resulting in economic gains. Analyzing the selection process, they show that information channels vary across forecasting horizons. Variable selection produces clusters and provides evidence that there are structural changes with regard to the significance of information channels.

The article provides unique quantitative insights about highly secretive and poorly understood speculation in the oil market. To demystify the behavior of speculators, the authors look at the problem from several different angles. First, they explain how the presence of the large over-the-counter (OTC) oil derivatives market leads to mischaracterization of traditional hedgers and speculators. They then explain what makes oil speculation special and different from speculation across many other commodities. Consequently, the authors identify the winners and the losers in the oil futures market. To model the behavior of the winners which they associate with fast-moving quantitative hedge funds, they develop a novel framework based on a simple neural-network algorithm. The authors conclude by analyzing a popular investment strategy of following the winners, or the hedge-funds’ flows.

This article notes that Artificial Intelligence (AI) is a tool, not a replacement. However, as it evolves, it will redefine our roles. Efficiency gaps will widen, the nature of work will change and standards of professionalism will rise. AI literacy is the way forward.

This article examines the risks associated with AI infrastructure investments, focusing on energy consumption, natural resource constraints, and historical parallels from past industrial buildouts. It explores factors that could lead to infrastructure obsolescence, market misalignment, and the ways that investment strategies may evolve in response to AI’s future.