查詢結果分析
相關文獻
頁籤選單縮合
題名 | 結合學習向量量化與協同過濾之交換混合式過濾電影推薦架構=Combining LVQ Collaboration Filtering on Switching Hybrid Movie Recommendation |
---|---|
作者姓名(中文) | 黃純敏; 林重佑; 黃進瑞; | 書刊名 | 資訊管理學報 |
卷期 | 20:4 2013.10[民102.10] |
頁次 | 頁423-447 |
分類號 | 312.2 |
關鍵詞 | 學習向量量化; 推薦系統; 混合過濾; 協同過濾; 內容過濾; Learning vector quantization; Recommendation system; Hybrid filtering; Collaborative filtering; Content-based filtering; |
語文 | 中文(Chinese) |
中文摘要 | 內容過濾與協同過濾是經常用於提供個人化服務的技術,近年來則多偏向結 合各種監督式學習的混合式過濾方式,並以三層或多層式網路架構產生推薦結 果,然而其設計不易且有網路收斂效率低的問題。本研究以學習向量量化(Learning Vector Quantization; LVQ)簡約的兩層式網路架構,運用交換(Switching)混合過 濾策略產生推薦內容。研究以MovieLens 資料集驗證方法架構。實驗發現,學習 向量量化可快速學習使用者多變的喜好。搭配交換混合過濾策略,可產生適切的 個人化內容,滿足不同使用者的推薦需求。研究結果顯示,本架構確可改善內容 過濾與協同過濾各自的缺點,整體精確率為79%,召回率為82%。 |
英文摘要 | Content-based filtering and collaborative filtering are often used to provide personalized services technology. Recently, lots of supervised neural networks are combined with hybrid recommendation and adopted three layers or multiple layers to construct recommendation. Their drawbacks are slow convergence and hard to design. In this paper, we presented a novel switching hybrid recommendation framework based on two-layer Learning Vector Quantization (LVQ) to provide personalized recommendations. MovieLens data set was used to test our framework and the experiment indicated LVQ can quickly detect and learn from user preferences. Results showed that switching hybrid strategy provides promising personalized recommendation and satisfied the needs of different users. Our experiment gains 79% of precision, and the recall rate also reaches 82%. |
本系統之摘要資訊系依該期刊論文摘要之資訊為主。