AI-Driven Restaurant Performance Optimization System Using Machine Learning and NLP for Predictive Management

Authors:

Mrs.S.B.V.Varalakshmi, Assistant Professor, Information Technology, MVGR College of Engineering. Vasamsetti Devi Neeharika, Computer science and Information Technology, MVGR College of Engineering, Vedula Keerthi, Computer Science and Information Technology, MVGR College of Engineering,

Vadarevu Sandeep, Computer Science and Information Technology, MVGR College of Engineering, Kujja Nithin, Computer Science and Information Technology, MVGR College of Engineering.

Abstract - Data-driven decision-making is a fundamental necessity in the competitive restaurant sector to maximize revenue and minimize operational waste. This project presents an integrated, deep learning and machine learning-based enterprise system capable of handling both quantitative transactional data and qualitative customer feedback. The system utilizes a multi-layered architecture: a Random Forest model enhanced with SHAP for transparent sales forecasting, the FP-Growth algorithm for market basket analysis, Logistic Regression for dynamic price elasticity simulations, K-Means clustering for RFM customer segmentation, and a BERT-based Natural Language Processing (NLP) module for sentiment analysis. The system is implemented using Python and deployed as an interactive Streamlit-based web application that allows managers to upload Point-of-Sale (POS) data and visualize financial projections. It processes historical data, identifies profitable synergies, and generates real-time financial waterfall metrics for pricing strategies. Experimental results demonstrate that the system achieves high predictive accuracy and operational flexibility compared to traditional management approaches. This work highlights the importance of Artificial Intelligence in real-world applications such as restaurant analytics and predictive business operations.

Keywords: restaurant optimization, machine learning, market basket analysis, price elasticity, BERT, computer vision, predictive management.