项目光谱：利用生成式人工智能强化通货膨胀临近预报

TitleTitleSubtitleEnabling climate risk analysis via Generative AIProject GaiaDecember 2023TitleTitleSubtitleProject SpectrumUsing generative AI to enchanceinflation nowcastingFebruary 2026 © Bank for International Settlements 2026.All rights reserved.Limited extracts may be reproduced or translated provided the source is stated. Executive summary The availability of web-scraped and scanner data sets provides central banks withunprecedented access to real-time data on individual product prices. However, to usethese data for inflation nowcasting and forecasting, analysts need to classify products Product classification at the scale of web-scraped data represents a major challenge.Manually processing this amount of data is not feasible. Classification using largelanguage models (LLMs) is promising, but with the LLM models currently available, theprocessing time and cost become prohibitively high. Project Spectrum used the EuropeanCentral Bank’s Daily Price Dataset (DPD), which contains billions of price-product daily Project Spectrum – a collaboration between the Bank for International Settlements(BIS), the Deutsche Bundesbank and the European Central Bank – explored an alternativeapproach where artificial intelligence (AI) was used only to transform product descriptionsinto high-dimensional text embeddings. These were then classified into product categoriesusing classic machine learning algorithms. Text embedding is a foundational AI technique Besides classifying all records in the current DPD, the project has developed aproduction pipeline solution that can classify new products as they are added to the DPD.In addition, to ensure continuous improvement, an iterative algorithm was implementedto gradually expand the reference data set. By selectively adding manually labelled data By turning raw, fragmented product descriptions into structured data, ProjectSpectrum equips analysts and policymakers with timely, detailed insights into pricedevelopments. Ultimately, the project contributes to an emerging new generation of AI- This report is intended for monetary policy analysts who utilise high-frequency datafor inflation nowcasting and data scientists within central banks looking for cost-effectivealternatives to LLMs for large-scale classification. It also serves as a technical reference forstatistical agencies seeking to automate the categorisation of scanner and web-scraped Table of contents Executive summary Acronyms and abbreviations4 1. Using online price data for inflation nowcasting 2. Related literature8 3. Spectrum overview9 4. Data4.1. Daily Price Data set (DPD) 124.2. Ground truth via manual labelling of reference and test data sets134.3. Sample representativeness15 5. Implementation17 5.1. Curated reference data set 5.2. Embedding model5.3. Classification using the k-nearest neighbour algorithm5.4. Classification using a feedforward neural network5.5. Direct large language model prompting 2222 6. Evaluation6.1. Classification accuracy 6.2. Feasibility and cost comparison of classification methods 7. Initial deployment and a continuous refinement process 8. Conclusions and next steps30 9. Project participants and Acknowledgements 10. Bibliography 1.Using online price data for inflation nowcasting Accurate and timely inflation nowcasts1and forecasts are central to effective monetarypolicy because inflation often responds to policy measures with a time lag.2By anticipatingfuture price trends, policymakers can adjust interest rates and other policy tools earlier tomaintain price stability and prevent costly economic fluctuations. Such foresight not only As a significant share of consumer spending is done online, data from e-commerceplatforms are a rich source of real-time information for inflation nowcasting and price-setting analysis. In traditional statistical sampling and in-store data collection, field officersvisit stores and document prices at periodic intervals. As online data capture real-timeprice variations, they have the potential to provide more timely and targeted insights thantraditional macroeconomic indicators, which are often available with some delay and at However, while online price data are rich in detail, extracting actionable insightsremains a challenge. The main issue is that web-scraped data are not standardised –they come in various formats, lack quality adjustments and contain key information in The principal challenge Project Spectrum addresses is the labelling of productsin accordance with international classification standards, such as the Classification ofIndividual Consumption According to Purpose (COICOP). To harness information onindividual prices for inflation analysis, it is crucial to map each product to a classification The advantages of improving the quality and accessibility of high-frequency price dataare clear. At the same time, the challenges of creating a structured, unified data set fromhigh-frequency data are manifold. In addi