Inflation prediction is essential for economic strategy and decision-making, but conventional techniques frequently underperform because they depend on lagging indicators and organized data such as the Consumer Price Index (CPI) and Wholesale Price Index (WPI). This study presents a combined method that merges Natural Language Processing (NLP) techniques and Large Language Models (LLMs) with structured data analysis to improve forecasting precision and adaptability. The hybrid model integrates organized economic indicators with unstructured data sources, such as news articles, social media contributions, policy declarations, and expert analyses. NLP methods, including sentiment analysis and topic modeling, derive insights from qualitative data to uncover trends, public sentiment, and new themes that impact inflation. These insights reveal immediate factors and hidden trends frequently missed by traditional approaches. LLMs are crucial in deciphering intricate economic dynamics due to their sophisticated contextual comprehension. They examine the interaction of elements such as fiscal and monetary policies, geopolitical events, and supply chain issues to offer detailed insights into inflationary patterns. This blend of structured and unstructured data analysis allows the model to predict inflation changes more precisely, even in unstable circumstances. The suggested method shows enhanced predictive accuracy in comparison to conventional techniques and offers practical insights for policymakers. By connecting structured and unstructured data, it provides a thorough insight into inflationary dynamics, facilitating more efficient and prompt decision-making. The combination of NLP and LLMs enhances forecasting accuracy while creating a strong foundation for tackling future economic issues in a world that relies more on data.
Pandey et al. (2025) studied this question.