SkyCity ProfitSim — Multi-Channel Restaurant Profit Prediction & Prescriptive Optimization System

SkyCity ProfitSim is an end-to-end, machine-learning-driven decision-support system designed to help restaurant and bar operators predict, simulate, and optimize profitability across multiple delivery and service channels — Uber Eats, DoorDash, and self-delivery — using a real-world dataset drawn from food-service establishments in Auckland, New Zealand. The system integrates three analytical layers: 1. Predictive Modelling: Multiple regression algorithms (Linear Regression, Random Forest, Gradient Boosting, and XGBoost) were benchmarked to predict net restaurant profit from operational and channel-level features. The tuned XGBoost model achieved the strongest performance with a test-set R² of 0.984. 2. Monte Carlo Simulation: A stochastic simulation engine propagates uncertainty in key cost and demand drivers through thousands of randomized trials, producing probabilistic profit distributions (rather than single-point estimates) to support risk-aware decision-making. 3. Prescriptive Optimization: A grid-search-based optimizer searches over controllable decision variables — including channel mix and pricing — to recommend the configuration expected to maximize net profit, subject to operational constraints. These capabilities are delivered through an interactive, multi-page Streamlit web dashboard, allowing non-technical users to explore historical analytics, generate predictions, run simulations, and view optimizer recommendations without writing code. This project was developed as a final-year B.Tech Computer Science and Engineering (Artificial Intelligence) capstone project at Gautam Buddha University, Greater Noida, India. Keywords: restaurant profitability, machine learning, XGBoost, Monte Carlo simulation, prescriptive optimization, decision-support system, Streamlit, food-delivery analytics.