Enhanced tourist flow forecasting in Aosta Valley: A novel ensemble AI framework with adaptive temporal dynamics

Area of Science:

• Artificial Intelligence
• Data Science
• Tourism Management

Background:

• Current tourism forecasting methods lack adaptability to environmental changes and struggle with sudden demand shifts.
• Existing approaches often fail to integrate multi-source data streams effectively.

Purpose of the Study:

• To develop an innovative ensemble artificial intelligence framework for monitoring and forecasting tourist flows.
• To address limitations in existing methods by introducing dynamic weighting and multi-source data integration.

Main Methods:

• Utilized a large dataset of over 41 million vehicle passages from 14 sensor portals.
• Developed the Adaptive Temporal Ensemble (ATE) algorithm, integrating XGBoost, Random Forest, Support Vector Regression, and LSTM with a meta-learning layer.
• Implemented dynamic weight adjustment based on temporal patterns, seasonality, meteorological conditions, and real-time performance.

Main Results:

• Achieved significant improvements in forecasting accuracy: 23.7% reduction in Mean Absolute Error (MAE) and 31.2% reduction in Mean Squared Error (MSE).
• The ensemble framework demonstrated R2 scores exceeding 0.94 for short-term predictions.
• Showcased robustness across different seasonal patterns and extreme weather conditions.

Conclusions:

• The adaptive ensemble AI framework offers superior accuracy and robustness for tourism flow forecasting.
• Provides practical benefits for destination management, including enhanced resource allocation and traffic management.
• Presents a scalable solution for intelligent tourism management applicable to other regions.