University of Twente Student Theses
As of Friday, 8 August 2025, the current Student Theses repository is no longer available for thesis uploads. A new Student Theses repository will be available starting Friday, 15 August 2025.
Large language models are successful retail forecasters
Laksberg, Andreas (2025) Large language models are successful retail forecasters.
|
PDF
864kB |
| Abstract: | Large language models (LLMs) have made headlines since the release of ChatGPT to the public in 2022. Quick advancements in their capabilities have made them a potential disruptor in many industries. This study investigated whether large language models could create and improve retail forecasting models without human interference and proposed a theoretical framework for adapting LLMs into retail forecasting. An experiment was conducted where two of the currently most advanced LLMs, OpenAI’s o4-mini-high and Gemini 2.5 Pro, were tasked with improving the accuracy of their forecast models over 10- and 20-attempt series based on Walmart sales data from the M5 forecasting competition. The results showed that ChatGPT beat the accuracy of the best performing benchmark model by 10.2%. Gemini outperformed most benchmarks but lost to the most accurate benchmark by 1.8%. Meanwhile, Gemini showed off its learning capabilities and achieved statistically significant improvements to accuracy over a series of attempts while ChatGPT failed to produce statistically significant improvements over time. This study has explored using LLMs in retail forecasting, highlighting the potential of LLMs being able to automate a significant amount of the forecasting process in the future. |
| Item Type: | Essay (Bachelor) |
| Faculty: | BMS: Behavioural, Management and Social Sciences |
| Subject: | 85 business administration, organizational science |
| Programme: | International Business Administration BSc (50952) |
| Link to this item: | https://purl.utwente.nl/essays/106768 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page