Deep Learning and IoT-Based Evaluation and Optimization for Tourism Accommodation Service Quality

Abstract

Traditional evaluation approaches, dependent on delayed feedback and manual inspection, fail to capture dynamic tourist requirements and operational fluctuations. To address these challenges, this research proposes an Adaptive Black Hole-Attention Bi-Directional Long Short-Term Memory (AdapBH-Att-Bi-LSTM) model that continuously senses, processes, analyzes, and improves accommodation service performance. The data capture layer integrates environmental, motion, RFID, occupancy, energy meter, and guest feedback streams to enable holistic monitoring. The analytics layer incorporates AdapBH Optimization for dynamic hyperparameter tuning, a soft attention mechanism for feature prioritization, and a Bi-LSTM to learn forward–backward temporal dependencies. The decision layer facilitates auto-scheduling, resource allocation, predictive maintenance, energy control, and comfort adjustment. Experimental evaluation demonstrates that the proposed framework achieves 0.989 prediction accuracy and significantly lower MAE (0.094) across all service types.