以社群為基礎的分散式非結構化對等網路上的內容搜尋：興趣適應逼近法

Abstract

在對等網路中，內容搜尋是最重要的問題之一。在這篇論文中，我們討論在一個特別的對等系統─以社群理論為基礎的對等系統中的內容搜尋。已經有些研究指出在以社群理論為基礎的對等系統中的內容搜尋可以更快速及更容易。我們發現在一個以社群理論為基礎的對等系統中的新問題─端點喜好的轉換。當端點們改變他們的喜好時，他們之前收集的知識就不再像過往一樣有用。因此我們提出一個非集中式興趣適應的方法來解決這個問題，一個興趣適應的內容搜尋（IACL）。我們提出的方法更有智慧的利用兩個在以社群理論為基礎的對等系統中的特質，「群聚」和「推薦」，以讓此方法可以去適應端點喜好的轉換。我們也做了些模擬實驗來說明在端點轉換喜好的環境中，我們的方法比其他的內容搜尋方法在某些實驗指標更好。從實驗數據中，我們得知我們提出的 IACL 有足夠的成功率搜尋到內容和在查詢中較低的訊息負擔。

Content location is one of the most important problems in peer-to-peer networks. In this thesis, we discuss content location in a special peer-to-peer network, the social-based peer-to-peer networks. There are some researches which show that locating content is faster and easier in social-based peer-to-peer networks. We discover a new problem in social-based peer-to-peer networks, peers change their tasty. While peers change their tasty, the knowledge they collected is not useful as past. Hence, we proposed a decentralized interest adaptive approach to solve it, an interest adaptive content locating (IACL). It makes the two characters of social-based “clustered” and “recommendation” more wisely; it adapts the behavior that peers change their tasty. We also do some simulation to show our approach is better than other content locating methods on some experiment indices in peers change their tasty environment. From the simulation results, we know that the IACL method we proposed has enough success rate to locate content and lower messages overhead while query.