人脈網絡的多重搜尋策略研究

Abstract

人脈網絡具有可搜尋的特性。藉由其他節點的幫助，我們通常可以只透過鄰近節點的資訊與互動來達成搜尋目的，搜尋過程通常為詢問鄰居一步步找尋答案可能的所在方向。在這樣的模式下，如何挑選合適的下一個傳送節點，成為我們想要探討的議題。此外，人脈網絡富涵人的資料，透過節點之間的互動，這些資料將可以被交換與分析，進而產生更有價值的資訊，考量這些資訊通常將有助於 提升搜尋的成效。 在本論文中，我們想要運用人身上所帶的資訊來幫助搜尋。藉由考量多維度資料的方式，我們可以在不同維度中挑選取出合適的候選人，之後再從這些候選人之中，進一步篩選出我們想要傳送的對象。透過這樣多層篩選的方式，我們最後的選擇傳送對象將可以結合有不同衡量方法的優點，進而提升搜尋成效。而因為節點可以依據他所擁有的資訊來選擇下一個傳送節點，也讓節點多了自主性與彈性。

Social network has the property that it is searchable. With the help from other nodes, we can search for some resource successfully only by utilizing neighbor's information and interaction between people. The process of searching is often achieved by asking neighbor step by step. Based on this approach, how to choose a appropriate next forwarding target becomes the topic we want to study. Besides, there are rich data about people in social network. By interaction of people, these data can be exchanged and analyzed to extract more information. Considering more people's characteristic or interaction data often can improve the performance of searching. In this thesis, we want to utilize the information on people when designing searching strategy. By considering multi-dimensional data, we can get some candidates after evaluating neighbors in different dimensional data. Then we can filter these candidates further. By this way, the filtering mechanism let us combine advantages of different evaluation process and then improve the performance of searching. This design also give the node some flexibility because node can decide the next forwarding target based on the information it has.