Advanced LSTM frameworks for hotel bookings leveraging hippopotamus-inspired optimization

The swift growth of Europe’s hospitality industry has resulted in an extensive selection of guest hotel options. The proliferation presents a challenge in choosing lodgings that optimally satisfy individual requirements, highlighting the necessity for advanced recommendation systems. This study employs the Hippopotamus Optimization Algorithm (HOA), a novel metaheuristic optimization technique, to tune the hyperparameters of Long Short-Term Memory (LSTM) networks, thereby enhancing the efficacy of these systems. These networks excel at analyzing substantial amounts of unstructured textual data from hotel reviews, deriving significant insights that indicate traveler preferences, attitudes, and experiences, hence supporting hotel booking likelihood prediction for decision support systems. The efficacy of LSTM networks significantly relies on the precise calibration of hyperparameters, including learning rate and hidden layer dimensions. The HOA, influenced by hippopotamuses’ eating and territorial behaviors, skillfully equilibrates the exploration and exploitation phases to determine ideal hyperparameter configurations. This methodology has been substantiated by an extensive examination of a significant dataset of European hotel reviews, whereby the HOA-optimized LSTM model exhibited a training loss of 28.30% with an accuracy of 97.69%, and a validation loss of 30.16% with an accuracy of 93.97%. This study introduces an innovative approach that markedly improves the provision of data-driven hotel booking prediction insights, advancing the domain of tourism technology.