A Model of E-commerce Recommender System using Enhancement Document Context Based on Attention and User Information Based on Auto Encoder and Latent Factor
Muh Hanafi · 2022 5th International Conference on Information and Communications Technology (ICOIACT) · 2022
Recommender system is essential tool in e-commerce application. Recommender system responsible to produce relevant information to customer. The successful adoption of recommender system influence of e-commerce company revenue. Collaborative filtering is the most successful of recommender system algorithm that adopted in many large e-commerce company. Collaborative filtering relies on rating matrix to produce product index recommendation. However, number of rating matrix very sparse due to majority customer are lazy to rate the product. In this research, we implemented product document information enhancement using attention mechanism that combined with stack denoising auto encoder (SDAE) to extract user information and latent factor based on probabilistic matrix factorization (PMF). According to our experiment report, our model success to improve the performance over previous work more than 18% in average over traditional PMF in ML.M and ML.10M.