隨著科技的進步與電商的蓬勃發展,用戶面對眾多的商品種類,無法藉由瀏覽搜索就得到想要的結果。現今的推薦系統(Recommendation System)廣泛應用於各行各業,可用於預測用戶的購買喜好。傳統的推薦系統只考慮到用戶以往購買過的商品以及用戶間購買商品的關係,而沒有考慮到用戶彼此的社交關係。然而用戶的購買行為往往受到同學、朋友、親戚和同事之影響。因此本論文提出使用深度學習(Deep Learning)的圖形神經網路(Graph Neural Network, GNN)的方法,利用用戶間親疏的社交關係與用戶購買商品的評分,加上購買商品間的關係,預測可能購買的商品,提升推薦商品的準確度。
With the advancement of technology and the vigorous development of e-commerce, users are faced with numerous product categories, and they cannot get the desired results by browsing and searching. The recommendation systems are widely used in various fields and can be used to predict users’ purchasing preferences. The traditional recommendation system only considers the products that users have purchased in the past and the relationship between the purchased products between users, but does not consider the social relations of users. However, users’ purchase behavior is often influenced by their classmates, friends, relatives and colleagues. Therefore, this thesis proposes a Graph Neural Network (GNN) mechanism based on deep learning that can predict the possible purchased products and get high accuracy of the recommended products from the social tie strengths.