過去幾年中，基於深度學習的模型（DL）受到了很多關注，尤其是在順序推薦任務領域。由於其處理複雜數據的能力，當前的順序DL研究工作已經超越了傳統模型，例如基於Markov鍊和基於因子分解的模型。但是，基於順序DL的模型的研究仍有改進的空間。特別是，如何設計有效的DL模型來處理不同場景下的順序推薦任務。在這種情況下，本文通過考慮現有方法的當前局限性，重點研究基於DL的順序推薦系統。具體來說，我們演示了順序推薦過程的概述概念，介紹了相關的最新算法，總結了影響基於DL的模型性能的關鍵因素，提出了用於復雜環境下順序推薦任務的新型基於DL的方法，並進行相應的評估，以顯示我們提出的模型在最新方法上的有效性。最後，我們通過系統地概述當前的挑戰，未來的方向以及我們在該領域的貢獻來結束我們的論文。最後，我們認為我們提出的項目對現有的序列感知推薦工作具有很高的積極貢獻。;Deep learning based models (DL) have received a lot of attention in the past few years especially with the domain of sequential recommendation tasks. Due to its capability to deal with complex data, currently sequential DL research works have surpassed traditional models such as Markov chain-based and factorization-based models. However, there is a room for improvement on the studies of Sequential DL-based models. Particularly, how to design an effective DL model to handle the sequential recommendation tasks under different scenarios. In this view, this thesis focuses on the DL-based sequential recommender systems by taking the current limitations of existing methods into consideration. Specifically, we demonstrate the overview concept of sequential recommendation processes, present the related state-of-the-art algorithms, summarize the key factors affecting the performance of DL-based models, propose the novel DL-based for sequential recommendation tasks under complex settings, and conduct corresponding evaluations to show the effectiveness of our proposed models over the state-of-the-art methods. We finally conclude our thesis by systematically outlining current challenges, future directions and our contributions in this field. At last, we believe that our proposed project has high positive contributions to the existing sequence-aware recommendation works.
