Singular value decomposition model application for e-commerce recommendation system : Aplikasi model dekomposisi nilai tunggal untuk sistem rekomendasi e-commerce

Wervyan Shalannanda (Institut Teknologi Bandung)
Rafi Falih Mulia (Institut Teknologi Bandung)
Arief Insanu Muttaqien (Institut Teknologi Bandung)
Naufal Rafi Hibatullah (Institut Teknologi Bandung)
Annisabelia Firdaus (Institut Teknologi Bandung)

JITEL (Jurnal Ilmiah Telekomunikasi, Elektronika, dan Listrik Tenaga)
Vol. 2 No. 2: September 2022

A recommendation system is one of the most important things in today’s technology. It can suggest products that match the user’s preferences. Many fields utilize this system, including e-commerce, using various algorithms. This paper used the matrix factorization-based algorithm, singular value decomposition (SVD), to make a recommendation system based on users’ similarities. Afterward, we implement the model against the ModCloth Amazon dataset. The results imply that the SVD algorithm yields the best accuracy compared to other matrix factorization-based algorithms with root mean square error (RMSE) of 1.055586. Then, we optimized the SVD algorithm by changing the hyperparameters of the algorithm to generate better accuracy and yield a model with an RMSE value of 1.041784.