Abstract
The paper presents a smart retail management system with the help of AI and the Retrieval-Augmented Generation (RAG) framework to facilitate highly flexible and accurate sales forecast. The integration of a Large Language Model (LLM) and vector database, which is supported by Snowflake, can make the system fast and efficient when it comes to searching data, but on the other hand, LlamaIndex is efficient in semantic searches and feature search of structured tables, which ensures deep conclusions on complex sets data. The research methodology is also based on a quite logical flow: the information is gathered using Kaggle, then it is shaped appropriately to allow missing values and the standardization of the inputs, and only the verified information is imported to Snowflake to be accessed to be able to make it quick. The system recognizes the appropriate sales trends, the customer behavioral trends, and the market indicators to produce data-driven and precise forecasting on the searches conducted with the cosine similarity. The performance measures were close in the evaluation of the model effectiveness where Mean Absolute error (MAE) was 95.7, Root Mean square error (RMSE) was 94.3, an r-square was 87 and F1-score was 94. These findings show that it is efficient in minimizing errors in prediction and trade-off possible accuracy and recall. Coupled with the Snowflake and its ability to search data successfully and LlamaIndex and its semantic extraction technique, it is possible to ensure the accuracy of the predictions is significantly higher and ensure the improvements in the management of inventory and resources. Furthermore, the hybrid strategy can be implemented to enhance the operational effectiveness, cost and agility reduction, as well as the sales strategy enhancement. Developments will be undertaken on future integration of outer market and social media information, means incorporation and exploration of higher hybrid architecture to enhance predictive accuracy, flexibility, and strength in retail landscapes of rapidly changing retail environments.
Showing the abstract — retrieve the full paper via the Exa API.