OPTIMASI MODEL LIGHT GRAPH CONVOLUTIONAL NETWORK PADA REKOMENDASI DESTINASI WISATA BERDASARKAN RIWAYAT INTERAKSI PENGGUNA

    Frahari Perdana Putra, - (2025) OPTIMASI MODEL LIGHT GRAPH CONVOLUTIONAL NETWORK PADA REKOMENDASI DESTINASI WISATA BERDASARKAN RIWAYAT INTERAKSI PENGGUNA. S1 thesis, Universitas Pendidikan Indonesia.

    Abstract

    Kemajuan pesat dalam jaringan digital berdampak pada lonjakan data di berbagai industri hiburan, termasuk pariwisata. Model rekomendasi menjadi solusi efektif untuk mengatasi masalah ini. Light Graph Convolutional Network (LightGCN), sebagai pendekatan berbasis Neural Network, LightGCN menawarkan keunggulan dalam menghubungkan antara pengguna dan destinasi wisata berdasarkan riwayat interaksi. Penelitian ini memiliki tujuan untuk mengevaluasi model LightGCN dalam rekomendasi destinasi wisata menggunakan dataset interaksi pengguna dan destinasi, serta melakukan analisis terkait pengaruh hyperparameter tuning, termasuk variasi konfigurasi dari embedding dimension, learning rate, lapis propagasi dan jenis optimizer terhadap kinerja model dalam memberikan rekomendasi yang relevan. Evaluasi dilakukan menggunakan metrik Normalized Discounted Cumulative Gain (NDCG) dan Recall. Hasil penelitian menunjukkan bahwa konfigurasi optimal dicapai dengan menggunakan optimizer Nadam, learning rate 0.005, embedding dimension 128, dan 5 lapisan propagasi. Model final yang dihasilkan menunjukkan performa yang sangat tinggi, dengan skor NDCG@5 sebesar 0.6341 dan Recall@5 sebesar 0.8011 pada test set. Hasil ini membuktikan bahwa proses hyperparameter tuning yang sistematis secara signifikan lebih unggul dibandingkan konfigurasi baseline dan mampu meningkatkan akurasi serta kualitas peringkat rekomendasi secara substansial. Menjadikan nya lebih efektif terhadap karakteristik dataset. Temuan penelitian ini memberikan wawasan praktis bagi pengembangan model rekomendasi dengan menetapkan sebuah benchmark performa yang kuat dan mengidentifikasi kombinasi parameter yang efektif untuk arsitektur LightGCN. ------- Rapid advances in digital networks have led to an explosion of data across entertainment industries, including tourism. Recommendation models offer an effective solution to this challenge. Light Graph Convolutional Network (LightGCN), as a neural‑network‑based approach, excel at linking users and tourist destinations based on their interaction histories. This study aims to evaluate the LightGCN model for recommending tourist destinations using a user-destination interaction dataset, and to analyze the impact of hyperparameter tuning specifically variations in embedding dimension, learning rate, number of propagation layers, and choice of optimizer on the model’s ability to deliver relevant recommendations. Evaluation is conducted using the Normalized Discounted Cumulative Gain (NDCG) and Recall metrics. Results show that the optimal configuration employs the Nadam optimizer, a learning rate of 0.005, an embedding dimension of 128, and five propagation layers. The final model achieves outstanding performance, with an NDCG@5 score of 0.6341 and a Recall@5 score of 0.8011 on the test set. These findings demonstrate that a systematic hyperparameter tuning process significantly outperforms the baseline configuration and substantially enhances both accuracy and ranking quality, making it highly effective for the characteristics of the dataset. The insights from this research establish a strong performance benchmark and identify an effective parameter combination for the LightGCN architecture.

    [thumbnail of S_RPL_2102972_Title.pdf] Text
    S_RPL_2102972_Title.pdf

    Download (1MB)
    [thumbnail of S_RPL_2102972_Chapter1.pdf] Text
    S_RPL_2102972_Chapter1.pdf

    Download (298kB)
    [thumbnail of S_RPL_2102972_Chapter2.pdf] Text
    S_RPL_2102972_Chapter2.pdf
    Restricted to Staf Perpustakaan

    Download (606kB) | Request a copy
    [thumbnail of S_RPL_2102972_Chapter3.pdf] Text
    S_RPL_2102972_Chapter3.pdf

    Download (552kB)
    [thumbnail of S_RPL_2102972_Chapter4.pdf] Text
    S_RPL_2102972_Chapter4.pdf
    Restricted to Staf Perpustakaan

    Download (779kB) | Request a copy
    [thumbnail of S_RPL_2102972_Chapter5.pdf] Text
    S_RPL_2102972_Chapter5.pdf

    Download (206kB)
    [thumbnail of S_RPL_2102972_Appendix.pdf] Text
    S_RPL_2102972_Appendix.pdf
    Restricted to Staf Perpustakaan

    Download (560kB) | Request a copy
    Official URL: https://repository.upi.edu
    Item Type: Thesis (S1)
    Additional Information: https://scholar.google.com/citations?hl=en&user=ENvxlGwAAAAJ&scilu=&scisig=ACUpqDcAAAAAaJMMB7ppMLD-SE-WmLISRUWqqn8&gmla=AH8HC4xB6CoL6cnof-nFzsVyVNlzFeqgwLar6rjY211X9Sggof5GgKqY6fAggdP1ETANIvgTpbPU5ShrQXYHYXpi69lDW1CfA0srXsc&sciund=13253882178658241475 ID SINTA Dosen Pembimbing: Yulia Retnowati: 6852573 Hendriyana: 6658557
    Uncontrolled Keywords: Hyperparameter Tuning, Interaksi, Light Graph Convolutional Networks, Normalized Discounted Cumulative Gain (NDCG), Rekomendasi Destinasi Wisata. Hyperparameter Tuning, Interaction, Light Graph Convolutional Networks, Normalized Discounted Cumulative Gain (NDCG), Tourist Destination Recommendation.
    Subjects: G Geography. Anthropology. Recreation > GV Recreation Leisure
    L Education > L Education (General)
    Q Science > Q Science (General)
    Q Science > QA Mathematics > QA75 Electronic computers. Computer science
    Q Science > QA Mathematics > QA76 Computer software
    T Technology > T Technology (General)
    Divisions: UPI Kampus cibiru > S1 Rekayasa Perangkaat Lunak
    Depositing User: Frahari Perdana Putra
    Date Deposited: 07 Aug 2025 03:17
    Last Modified: 07 Aug 2025 03:17
    URI: http://repository.upi.edu/id/eprint/135216

    Actions (login required)

    View Item View Item