Tourismai: An AI-Driven Tourism Demand Forecasting and Overcrowding Prediction Platform using Machine Learning

Devam Dilipbhai Patel                            Ms. Kiran Sharma                                Ms. Twinkle Shah

Dept. of Computer Engineering, PIT      Faculty Mentor, Asst. Professor, PIT         Industry Mentor, Infolabz IT services

Parul University, Vadodara, Gujarat.      Parul University, Vadodara,Gujarat,                       Ahmedabad, Gujarat

devampatel2004@gmail.com

Abstract— India's tourism sector contributes nearly 9.2% to the national GDP and employs upward of 87 million people; yet the infrastructure surrounding demand prediction, crowd management, and travel planning remains fragmented and largely manual. This paper introduces TourismAI, a full-stack, data-driven web platform developed in Python and Streamlit that attacks this problem from three complementary directions. First, a Random Forest Regressor trained on a curated 20,000-record tourism dataset delivers visitor-count forecasts with an R² score of 0.87 and a Mean Absolute Error of approximately 1,820 visitors per destination-date query. Second, a hand-crafted Travel Intelligence Engine (TIE) translates seasonal and destination-specific domain knowledge into interpretable overcrowding risk scores — High, Medium, or Low — achieving 91% classification accuracy across 50 verified test cases. Third, a knowledge-base-driven AI chatbot handles travel guidance queries through a hybrid keyword-matching and dynamic data-aggregation strategy. Beyond these core intelligence modules, TourismAI incorporates a real-time Business Intelligence dashboard with five interactive Plotly charts, a role-based access system distinguishing Travel Agents from regular Users, an interactive Map Explorer, a Budget Trip Planner with day-wise itinerary generation, a Place Comparison Engine, and a high-risk Alerts module. The modular architecture is built for extensibility: swapping in a new dataset or adding a new destination to the intelligence engine requires changes in only one file. TourismAI demonstrates how a relatively small, well-curated dataset combined with thoughtful ML and rule-based design can produce a genuinely useful decision-support tool for tourism stakeholders across India.

Keywords — Tourism Demand Forecasting, Machine Learning, Random Forest Regressor, Overcrowding Prediction, Streamlit, Python, Business Intelligence, AI Chatbot, Seasonal Risk Classification, Smart Tourism, India Tourism, Data-Driven Planning, Scikit-learn, Plotly, Role-Based Access Control