隨著電子商務和手機的普及，推薦系統就扮演著重要角色來幫助使用者找到他所喜愛的商品。然而，推薦系統面臨兩個問題，數據稀疏問題和使用者冷啟動問題。其中有一些研究使用跨域推薦來解決數據稀疏問題，將信息從豐富的領域轉移到稀疏的領域。在本文中，我們開發了一個新的跨域推薦系統框架(CD-ELR)，通過結合 MF 和循環神經網絡 (RNN)。首先，我們透過潛在因子模型 (MF) 學習用戶和項目的潛在因子。然後，我們選擇兩個組合算子，即Max-pooling和Average-pooling，來組合共同用戶的潛在因子。在獲得目標域和來源域中共同用戶的潛在因子後，我們使用循環神經網絡（RNN）捕捉潛在因子的演變以預測未來的用戶偏好。實驗結果顯示，CD-ELR 的性能優於其他的推薦演算法。此外，我們也在幾個真實世界的數據集上進行了實驗，以證明所提出的 CD-ELR 的實用性。;With the popularity of e-commerce and mobile phones, recommender system becomes a necessary tool to help users find their desired commodities. However, recommendation systems suffer two critical issues, sparsity problem and cold start problem. Several studies use cross domain recommendation to address the data sparsity problems, which transfer the information from the richer domain to the sparser domain. In this paper, by combining the MF and recurrent neural network (RNN), we develop a new framework for cross-domain recommender system namely CD-ELR. At the first step, we aim to learn the latent factors of users and items by latent factor model, MF. Then, we choose two combination operators, i.e., Max-pooling, and Average-pooling, to combine the latent factors of common users. After obtain the latent factors of common users in target and source domain, we capture the evolutions of latent factors by recurrent neural network (RNN) to make prediction of user preference in the future. The experimental results show that CD-ELR has better performance than other state-of-the-art recommendation systems. In addition, we carry out the experiments on several datasets to prove the practicability of proposed CD-ELR.
