Artificial Intelligence-Driven Prediction of Hotel Booking Demand for Revenue Optimization

Abstract

The rising competitiveness in the hospitality industry requires proper and smart forecasting models to maximize the hotel booking demand and revenue management plans. The proposed study is a hybrid CNN-Bi-LSTM model to predict the canceling of hotel bookings based on the Kaggle hotel booking demand dataset. An organized preprocessing system adopted, such as the missing value imputation, the treatment of ADR outliers, the encoding of labels, feature engineering, correlation-based feature importance analysis and, to enhance the model's dependability and data quality, the class balancing performed using SMOTE. The architecture proposed combines Convolutional Neural Networks (CNN) to obtain meaningful feature representations and Bidirectional Long Short-Term Memory (Bi-LSTM) networks to learn temporal patterns of booking. The applied approach has proven to be effective based on the results of performance evaluation with the help of confusion matrix-based measures 98.95% accuracy (acc), 98.82% precision (prec), 99.91% recall (rec), and 98.90% F1-score (F1). A comparative analysis indicates that the hybrid CNN-Bi-LSTM is much better than the classical models, which include Decision Tree, Random Forest, and standalone XGBoost. The results affirm the soundness, generalization and applicability of the suggested framework regarding the intelligent hotel booking demand forecasting and decision-making