Enhance Document Contextual using Attention-LSTM to Handle Sparse Rating Matrix For E-Commerce Recommender System

Hanafi, Hanafi (2022) Enhance Document Contextual using Attention-LSTM to Handle Sparse Rating Matrix For E-Commerce Recommender System. [Research]

Text (Research) ICAITI 2022 Submission 145.pdf Download (179kB)

Abstract

E-commerce is the most important service in last 2 decade. E-commerce service influnce growth of economy impact in world wide. Recommender system is essential mechanism to calculate product information for e-commerce user. The successfulness in recommender system adoption influence target revenue of e-commerce company. Collaborative filtering (CF) is the most popular algorithm to create recommender system. CF applied matrix factorization mechanism to calculate relathionship between user and product using rating variable as intersection value between user and product. However, number of rating very sparse where the number of rating only less than 4%. Product Document is the product side information representation. The document aims to advance matrix factorization work. This research consider to enhancement document context using LSTM with attention mechanism to capture contextual understanding of product review and incorporate with matrix factorization based on probabilistic matrix factorization (PMF) to produce rating prediction. This study employ real dataset using MovieLens dataset ML.1M, ML.10M and IMDB to observed our model called ATT-PMF. Movielens dataset represent of number sparse rating that only contains below 4% (ML.1M) and below 1% (ML.10M). Our experiment report show that ATT-PMF outperform than previous work morethan 2% in average. Moreover, our model also suitable to implement on huge datasets. For ICAITI 2022 (author) further research, enhancement of product document context will promising factor to handle sparse data problem in big data issue.

Actions (login required)

View Item