Implementasi Rekomendasi Produk E-Commerce Menggunakan Metode Matrix Factorization Fusing Reviews (MFFR)

Buaton, Ronaria Elisabeth (2025) Implementasi Rekomendasi Produk E-Commerce Menggunakan Metode Matrix Factorization Fusing Reviews (MFFR). Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5026211047-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (3MB) | Request a copy

Abstract

Sistem rekomendasi merupakan salah satu komponen berpengaruh dalam e-commerce untuk meningkatkan pengalaman pengguna dan konversi penjualan. Namun, metode tradisional seperti Matrix Factorization (MF) menghadapi tantangan kelangkaan data yang mengurangi akurasi prediksi preferensi pengguna. Tugas akhir ini mengusulkan model Matrix Factorization Fusing Reviews (MFFR) yang mengintegrasikan data ulasan dan rating pengguna menggunakan pendekatan analisis topik berbasis Latent Dirichlet Allocation (LDA). Metodologi mencakup tiga tahapan utama: (1) ekstraksi topik dari ulasan menggunakan LDA, (2) penggabungan matriks preferensi ulasan dan matriks rating melalui latent-factor matrix shared representation, dan (3) evaluasi model menggunakan metrik MAE, RMSE, Hit Ratio (HR), dan Normalized Discounted Cumulative Gain (NDCG). Tugas akhir ini menggunakan Amazon Beauty Set data yang memiliki karakteristik kelangkaan data yang tinggi. Hasil yang diharapkan adalah model rekomendasi yang lebih akurat dan personal dalam mengatasi masalah kelangkaan data pada platform e-commerce, khususnya di kategori produk kecantikan.
======================================================================================================================================
Recommender systems are essential components of e-commerce platforms, aimed at enhancing user experience and increasing sales conversions. However, traditional approaches such as Matrix Factorization (MF) often struggle with data sparsity, which significantly hinders their ability to accurately predict user preferences. This final project presents a model called Matrix Factorization Fusing Reviews (MFFR), which integrates user reviews and rating data by leveraging topic modeling through Latent Dirichlet Allocation (LDA). The implemented methodology consists of three main stages: (1) extracting latent topics from user reviews using LDA, (2) combining the review-based preference matrix and the rating matrix using a shared latent factor representation, and (3) evaluating the model’s performance using metrics such as MAE, RMSE, Hit Ratio (HR), and Normalized Discounted Cumulative Gain (NDCG). The model was applied to the Amazon Beauty dataset, which is known for its high level of data sparsity. The results demonstrate that the MFFR approach produces more accurate and personalized product recommendations, effectively mitigating the sparsity issue—particularly in the context of beauty product recommendations on e-commerce platforms.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Sistem Rekomendasi, Matrix Factorization, Latent Dirichlet Allocation, Ulasan Pengguna, Kelangkaan Data. Recommender System, Matrix Factorization, Latent Dirichlet Allocation, User Reviews, Data Sparsity.
Subjects:	T Technology > T Technology (General) > T57.8 Nonlinear programming. Support vector machine. Wavelets. Hidden Markov models. T Technology > T Technology (General) > T58.6 Management information systems
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Information System > 57201-(S1) Undergraduate Thesis
Depositing User:	Ronaria Elisabeth Buaton
Date Deposited:	26 Jul 2025 07:50
Last Modified:	26 Jul 2025 07:50
URI:	http://repository.its.ac.id/id/eprint/122285

Actions (login required)

View Item