Cultural Tourism Recommender Based on User Behaviour Modelling and Polynomial‑Based Graph Convolutional Neural Networks for Personalized Experiences

Abstract

Cultural tourism has witnessed a surge in popularity as travelers seek unique and personalized experiences that cater to their individual preferences. Cultural tourism presents challenges due to diverse user preferences and evolving interests. Traditional recommender systems often struggle to adapt to these dynamics, necessitating a more sophisticated approach. Hence, the limitations in adaptability observed in traditional approaches underscore the need for an innovative solution. In this manuscript, Polynomial‑Based Graph Convolutional Neural Networks (PGCNN) is proposed. Initially data is taken from TRD dataset. Afterward the data is fed to federated neural collaborative filtering (FedNCF) based pre-processing process. The pre-processing output is given to Adaptive and Concise Empirical Wavelet Transform (ACEWT) to extract the optimal features for enhancing the discriminative power and capturing intricate patterns in the cultural tourism data, thereby contributing to the improved accuracy and effectiveness of the recommendation system. After that, the extracted features are provided to Polynomial‑Based Graph Convolutional Neural Networks (PGCNN). The PGCNN is employed to enhance the accuracy of personalized cultural tourism recommendations. The PGCNN is used to model intricate relationships within user behavior data, capturing nuanced preferences and historical interactions. By leveraging the intelligent capabilities of the neural network, the objective is to provide accurate and context-aware recommendations, adapting dynamically to individual users' evolving cultural interests. The learnable parameters of the PGCNN is optimized using LOA. MATLAB is used to implement the proposed approach, and the proposed method's effectiveness To estimate CTR-UBM-PGCNN, several performance assessment criteria are employed, such as recall, accuracy, precision, f1-score, and computational time. The proposed CTR-UBM-PGCNN method shows the highest accuracy of 99%, precision of 99%, and F1-score of 97% while comparing other existing methods such as Cultural Tourism Recommender Based On User Behaviour Modelling using Convolutional Neural Network (CTR-UBM-CNN), Cultural Tourism Recommender Based On User Behaviour Modelling using Deep Learning (CTR-UBM-DL) and Cultural Tourism Recommender Based On User Behaviour Modelling using Machine Learning (CTR-UBM-ML respectively.