近年來隨著網路的蓬勃發展,網路上的遠距離互動已經不僅僅是使用者與使用者的對話,已經發展成由機器自動回應使用者的地步,也就是所謂的人工智慧,而近年來很火紅的深度學習演算法,在人工智慧的領域,不論是語音處理、電腦視覺與自然語言的處理等,領域都取得非常大的成就。相對來說,深度學習在推薦系統的領域上還是處於一個早期的探索階段。因此本論文提出利用深度學習的推薦系統,此種方式可以有效解決以往基於內容推薦的系統中常遇到的問題,像是冷啟動的問題等,而且有效的利用Alternating Least Squares(ALS)將用戶(User)對商品(Item)的評分兩個矩陣,充分的將數據中大量的缺失項目補足且減少矩陣維度,再利用Collaborative Filtering協同過濾技術找出用戶有相似行為的群體訊息,並根據這些訊息給用戶推薦。本論文中推薦系統能有效的解決傳統推薦系統相對不足的部分,並利用深度學習的方式,更準確的推薦用戶可能喜歡的商品。 The recommendation system is major top of discussion in the E-commerce , and Deep Learning algorithm became more than more popular , no matter Artificial Intelligence , Voice processing or Natural Language Processing , Deep Learning can be applied very success. The recommendation system field for Deep Learning is early to development . so we propose our methods , Deep learning Recommendation system with collaborative filtering , our method can solve many problems from the old recommendation system , like Dataset problem , Decision tree. And we using alternating least squares to find the item of user the score vector , and Collaborative Filtering to find the group message of similarly , recommend for the user according to the message.