摘要
隨著商業上的需求漸增，推薦系統在近年來變得越來越重要的議題。推薦系統提供使用者個人化的推薦，有許多的演算法在過去的很多論文中被提出，尤其是 Collaborative Filtering (CF) 類演算法提供了個人化推薦很高的精準度，但是這些論文中影音資料來源大多來自於Netflix Prize和Movielens等特定的Dataset，為了避免推薦演算法只適用於特定Dataset，本篇論文將實作推薦系統在真正的影音系統上，利用真實收集到的資料來驗證 CF 演算法。然而在推薦系統的討論也不僅在於精準度而已，過高的精準度通常也會使推薦結果不如預期的好，本篇論文更會討論推薦系統的多樣性，在個人化推薦曝光給使用者後，驗證演算法的精準度及多樣性。
並列摘要
Recommendation systems are becoming an increasingly important issue in recent years and providing personalized recommendations to individual users. Many algorithms have been proposed in many papers. In particular, the Collaborative Filtering (CF) algorithms provides high accuracy for personalized recommendations, but data sources come from specific datasets such as Netflix Prize and Movielens. In order to avoid the recommendation algorithm only applicable to specific datasets, this paper will implement the recommendation system on the video system in the real world, using the collected data to verify the ability of CF algorithms. However, the discussion of the recommendation system is not only accuracy. Actually, too high accuracy of the recommendation result is not as good as expected. Therefore, we will discuss the diversity of the recommendation system too.