Shipping's AI Forecasters Confront the Limits of Historical Data
Key Takeaways
- •A review of 28 studies from 2012 to 2024 found growing reliance on vessel supply data, commodity demand indicators, bunker fuel prices, and economic metrics to predict freight markets.
- •Geopolitical shocks, regulatory changes, and sudden supply disruptions cannot be reliably captured from historical data alone, limiting even the most advanced forecasting models.
- •Industry professionals agree that AI should function as a decision-support tool rather than a replacement for seasoned commercial judgment in shipping.
- •Experts warn that if market participants adopt similar models and datasets, herd behavior could intensify and accelerate market movements beyond what fundamentals justify.
- •Roar Adland of SSY argues that large language models add little value over existing machine-learning systems and remain unsuitable for autonomous commercial decisions.
Machine learning is steadily improving freight-rate forecasting in the maritime shipping industry, yet the sector's inherent volatility continues to elude even the most advanced models.
A review of 28 studies published between 2012 and 2024 found a growing reliance on vessel supply data, commodity demand indicators, bunker fuel prices, and broader economic metrics to predict freight markets. However, the review also concluded that geopolitical shocks, regulatory changes, and sudden supply disruptions cannot be reliably captured from historical data alone, underscoring a core limitation for an industry where rates can move quickly on events outside normal trend lines.
Industry figures broadly agree that artificial intelligence is best treated as a tool for decision support rather than a replacement for human judgement. Burak Cetinok of Arrow told sister title SplashTech that modern tools can process datasets faster than ever, but weak data quality and limited coverage still constrain accuracy. Panos Patsadas of Trans Global Projects argued that shipping remains insufficiently digitised and cautioned that past cycles are poor guides to future market behaviour, a reminder that forecasting remains tied to the quality and completeness of the underlying market record.
Christoffer Svärd of Sea noted that model output depends heavily on configuration, prompting, context, and human interpretation.
Roar Adland of SSY struck a more sceptical tone, arguing that large language models add little to the machine-learning systems that have already been in use for more than a decade and remain unsuitable for autonomous commercial decisions.
Experts identified three principal drivers of freight rates: predictable fundamentals, unpredictable shocks, and short-term herd behaviour. AI may sharpen analysis of the first category, but it offers little leverage on the second and could even amplify the third. If market participants rely on similar models and datasets, they may crowd into the same positions, accelerating market moves and undermining the very forecasts those tools were designed to produce.
The consensus among industry professionals is that AI can refine analysis, reveal hidden patterns, and improve timing, but it will not eliminate shipping cycles or substitute for seasoned commercial judgement.
Source: Splash247