Design and Implementation of an AI‐Driven Intelligent Decision Support System for Sustainable Ecotourism Development

Abstract

The development of sustainable ecotourism requires intelligent analytical frameworks to support sustainable tourism growth while preserving natural resources. The advancement of Artificial Intelligence (AI) and Machine Learning (ML) technologies enables organizations to manage resources through data‐driven analysis, but existing models do not successfully model the complex connections that exist between tourism patterns and environmental impact and resource consumption. Existing systems face challenges in capturing nonlinear relationships, providing interpretable outcomes, and accurately identifying ecotourism risk levels. This study introduces a hybrid AI framework that uses TabNet and XGBoost to predict ecotourism risks through its multi‐class prediction system. The model utilizes a dataset containing visitor numbers, resource consumption, visitor satisfaction, temperature, Air Quality Index (AQI), noise levels, timestamps, and location information. The preprocessing steps include data imputation, normalization, temporal feature extraction, and the generation of composite indicators, including the Tourism Pressure Index (TPI) and Environmental Stress Score (ESS). TabNet employs an attention mechanism to extract significant feature representations, whereas XGBoost classifies samples into three risk categories: low, medium, and high. The experimental results demonstrate that the proposed framework achieves an accuracy of 96.5% with improved precision, recall, F1‐score, and ROC‐AUC performance. The proposed framework enhances prediction accuracy while providing interpretable feature importance analysis for effective ecotourism risk assessment.