AI BriefingLG AI Research 646
Key point
LG AI Research has developed a mid-to-long-term forecasting model based on a Causal Map that structures the causal relationships of lithium prices using an LLM.
Details
Aiming for 'trustworthy forecasting' that serves as a basis for decision-making beyond mere accuracy, LG AI Research and LG Energy Solution collaborated on a mid-to-long-term lithium price forecasting project.
Lithium is a key raw material for secondary batteries, and due to its low supply elasticity, price volatility is very high when there is a supply-demand imbalance. In particular, since Upstream (mine and salt lake) investment requires a lead time of several years, structural mid-to-long-term forecasting, rather than simple time-series forecasting, is essential.
Existing deep learning models are strong at learning past patterns but have limitations in explaining complex market factors. To address this, the project focused on structuring the relationships among the key factors driving lithium prices.
Using LLM-based Causality Extraction techniques, cause-and-effect relationships were extracted from unstructured text data such as market outlook reports and news. Through this, a Causal Map was built showing how various variables—including demand, supply, investment, and policy—contribute to price formation.
This approach goes beyond simple numerical prediction by explaining the market structure and assumptions from which the forecast values are derived, providing trustworthy grounds for decision-makers who deploy large amounts of capital.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.