| DC 欄位 |
值 |
語言 |
| DC.contributor | 資訊管理學系 | zh_TW |
| DC.creator | 張博雅 | zh_TW |
| DC.creator | Po-Ya Chang | en_US |
| dc.date.accessioned | 2022-7-6T07:39:07Z | |
| dc.date.available | 2022-7-6T07:39:07Z | |
| dc.date.issued | 2022 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:444/thesis/view_etd.asp?URN=109423022 | |
| dc.contributor.department | 資訊管理學系 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 隨著行動應用程式的數量急遽成長，琳琅滿目的行動應用程式使得用戶要找到適合的選擇是困難且費時的。因此，幫助用戶解決此問題的行動應用程式推薦成為當務之急。過去研究解決此一問題常見的兩種做法，一種是對評分資料實施矩陣分解，以預測目標用戶對其他應用程式的可能評分。第二種做法則是根據用戶評論文本去預測評分。
本研究除了結合以上兩種傳統的作法外，更結合了深度學習的模式以產生更佳的模型。對於評分數據，我們採用深度矩陣分解來取得對應嵌入，以此增進用戶與項目的向量表示；對於評論文本，我們結合了多種嵌入方法所生成的表示以豐富輸入資訊，也更進一步提升評論文本所帶來的優勢。此外，我們的方法也結合了具備非線性轉換能力的深度神經網路，以提取更具代表性的抽象特徵。
我們透過蒐集自真實世界的數據集進行實驗，大量的實驗結果表明，我們提出的方法性能優於其他基線方法。另外，我們透過廣泛的實驗評估不同實驗設置的性能，同時也證明了我們所使用的多嵌入方法可以提供更好的推薦性能。 | zh_TW |
| dc.description.abstract | With the explosion in the number of mobile applications, it becomes difficult and time-consuming for users to find the most suitable and interesting apps from the millions of applications that exist today. Therefore, mobile application recommendation has become an immediate priority to help people address this problem. There are two common approaches to this question in previous studies, one is applying matrix factorization to the rating data to predict ratings of other applications by the target users. Another approach is to predict ratings based on users’ reviewing text. In addition to combining the above two traditional approaches, this study also incorporates deep learning methods to generate better models. For rating data, we use the deep matrix factorization method to obtain the latent factor vectors of users and items from the rating matrix as the input of the rating learning module to improve the representation of the rating. For the review text, we improved the text representation by using a multi-embedding approach. On top of that, our model also combines with the deep neural network which has the nonlinear transformation ability to extract more representative abstract features. We conduct experiments on the dataset which is collected from the real world. A large number of experimental results show that our proposed method has better performance than other existing methods. In addition, we evaluate the performance of different experimental setups through extensive experiments and proves that the multi-embedding approach we used can improve recommendation performance. | en_US |
| DC.subject | 推薦系統 | zh_TW |
| DC.subject | 深度學習 | zh_TW |
| DC.subject | 神經網路 | zh_TW |
| DC.subject | 矩陣分解 | zh_TW |
| DC.subject | 多嵌入 | zh_TW |
| DC.subject | Recommender system | en_US |
| DC.subject | Deep learning | en_US |
| DC.subject | Neural networks | en_US |
| DC.subject | Matrix factorization | en_US |
| DC.subject | Multi-embedding | en_US |
| DC.title | A Deep Learning Based Android Apps Recommendation with Multi-embedding and Matrix Factorization | en_US |
| dc.language.iso | en_US | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |