電機資訊學院: 資訊工程學研究所指導教授: 鄭卜壬施亭屹Shih, Ting-YiTing-YiShih2017-03-032018-07-052017-03-032018-07-052015http://ntur.lib.ntu.edu.tw//handle/246246/275453當新商品 (New item) 的數量成長速度越來越快,推薦系統 (Rec- ommendation system) 就越難兼顧到每一個新商品的曝光度 (Exposure)。 因此,我們提出了一套兩段式的推薦方法,期望能幫助新上架商 品增強曝光的機會。本篇論文所提出的方法有別於以往的協同過濾 (Collaborative Filtering) 推薦,在推薦商品時不僅僅考慮使用者的滿意 度或是商品的品質,也將品質未知的新上架商品推薦給可能願意提供 評價的使用者。我們可透過蒐集得到的評價確認新商品的品質,再決 定是否繼續推廣或抑制新商品。如此一來,我們僅犧牲了一點使用者 收到滿意商品的穩定性,卻換取了所有新上架商品極需的曝光度,讓 他們都有相同的機會被看見。我們的實驗實施在現有的 MovieLens 和 Netflix 資料上,而結果顯示了此種推薦方法的可行性。New items, e.g., mobile apps and movies, have been growing so fast that most of them cannot get discovered in a recommendation system. We propose a two-stage approach to appropriately promote new items. Different from pre- vious works on Collaborative Filtering (CF), our approach is not based only on item quality or user satisfaction. We force the new items to be promoted to those who would be potentially able to give ratings, and then leverage the gathered user preference to punish the promoted items with low quality in- trinsically. By slightly sacrificing the benefit of recommending the best items in terms of item quality or user satisfaction, our solution seeks to provide all of the items with a chance to be visible equally. The result of the experiments conducted on MovieLens and Netflix data demonstrates the feasibility of the approach.1032182 bytesapplication/pdf論文公開時間: 2020/8/21論文使用權限: 同意有償授權(權利金給回饋學校)推薦推薦系統商品曝光度曝光度協同過濾冷開始recommendationrecommendation systemitem exposureexposurecollaborative filteringcold startthesishttp://ntur.lib.ntu.edu.tw/bitstream/246246/275453/1/ntu-104-R02922038-1.pdf