Implementasi Model Self Attentive Sequential Recommendation Untuk Sistem Rekomendasi Pada Data Ulasan Produk E Commerce

Rachmad Santoso, Marro Teguh (2025) Implementasi Model Self Attentive Sequential Recommendation Untuk Sistem Rekomendasi Pada Data Ulasan Produk E Commerce. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 06111940000032-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Perkembangan industri e-commerce telah menghasilkan volume data interaksi pengguna yang sangat besar, seiring meningkatnya aktivitas belanja secara daring. Dalam situasi ini, sistem rekomendasi menjadi komponen penting untuk membantu pengguna menemukan produk yang relevan serta meningkatkan efektivitas pemasaran. Namun, banyak pendekatan yang ada belum sepenuhnya memanfaatkan informasi urutan interaksi pengguna secara optimal, sehingga hasil
rekomendasi masih dapat ditingkatkan. Penelitian ini mengimplementasikan model Self Attentive Sequential Recommendation (SASRec), yaitu model machine learning berbasis mekanisme self-attention yang dirancang untuk memahami pola perilaku pengguna dari urutan historis interaksi produk. Model ini dieksplorasi dan ditingkatkan melalui penyesuaian arsitektur dan parameter pelatihan guna menghasilkan prediksi produk yang lebih akurat. Data interaksi pengguna diproses melalui beberapa tahapan, termasuk pemetaan ID pengguna dan produk, pemisahan data menggunakan metode leave-one-out, padding, serta penerapan teknik sliding window untuk menghasilkan urutan yang seragam. Model dibangun dengan embedding layer, self-attention layer, dan feed-forward layer, serta dilatih untuk memprediksi produk yang kemungkinan besar diminati pengguna selanjutnya. Evaluasi dilakukan menggunakan dua metrik utama, yaitu Hit Rate (HR) dan Normalized Discounted Cumulative Gain (NDCG). Hasil eksperimen menunjukkan bahwa model SASRec yang diimplementasikan berhasil mencapai skor HR sebesar 0,6281 dan NDCG sebesar 0, 4811. Hasil ini menunjukkan bahwa model mampu menghasilkan rekomendasi dengan tingkat relevansi dan kualitas peringkat yang cukup baik. Temuan ini mengindikasikan bahwa pendekatan berbasis self-attention berpotensi untuk terus dikembangkan dalam membangun sistem rekomendasi e-commerce yang lebih adaptif, personal, dan efektif dalam memahami preferensi pengguna.
==========================================================================================================================================
The rapid growth of the e-commerce industry has led to a massive volume of user interaction data, driven by the increasing popularity of online shopping. In this context, recommender systems have become a crucial component to help users discover relevant products and enhance marketing effectiveness. However, many existing approaches have yet to fully utilize the sequential nature of user interactions, resulting in recommendations that are less contextual and personalized. This study implements the Self-Attentive Sequential Recommendation (SASRec) model, a machine learning approach based on self-attention mechanisms designed to capture user behavior patterns from historical product interaction sequences. The model is explored and improved through architectural adjustments and training parameter optimization to produce more accurate product predictions. User interaction data undergo several preprocessing steps, including user and item ID mapping, data splitting using the leave-one-out method, padding, and the application of a sliding window technique to generate uniform sequences. The model is constructed with embedding layers, self-attention layers, and feed-forward layers, and is trained to predict the next product most likely to be of interest to the user. The evaluation is conducted using two main metrics: Hit Rate (HR) and Normalized Discounted Cumulative Gain (NDCG). Experimental results show that the implemented SASRec model achieves an HR score of 0.6281 and an NDCG score of 0.4811. These results indicate that the model is capable of generating recommendations with a good level of relevance and ranking quality. These findings suggest that a self-attention-based approach has strong potential for further development in building e-commerce recommender systems that are more adaptive, personalized, and effective in understanding user preferences

Item Type:	Thesis (Other)
Uncontrolled Keywords:	SASRec (Self-Attentive Sequential Recommendation, self-attention, Hit Rate (HR), Normalized Discounted Cumulative Gain (NDCG). SASRec (Self-Attentive Sequential Recommendation, self-attention, Hit Rate (HR), Normalized Discounted Cumulative Gain (NDCG).
Subjects:	Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science)
Divisions:	Faculty of Mathematics and Science > Mathematics > 44201-(S1) Undergraduate Thesis
Depositing User:	Marro Teguh Rachmad Santoso
Date Deposited:	05 Aug 2025 03:49
Last Modified:	05 Aug 2025 03:49
URI:	http://repository.its.ac.id/id/eprint/127264

Actions (login required)

View Item