隨著電子化的趨勢，各式各樣的電子資料一直在網路上迅速的繁衍與增長。雖然已有相當多研究投入文件資料搜尋，但目前似乎面臨成果不佳以及發展膠著的窘境。因此本研究致力於發展正確性、實用性及方便性的演算步驟以提昇檢索成效。 在眾多的相關研究議題中，群聚經常被利用於大型資料的歸類，但在搜尋引擎的檢索中群聚研究並未多見，故本論文將以此方向做為研究重心。我們利用關聯規則（Association Rule）挖掘共通屬性之間的關係，再藉由圖解模型（Graph Model）的架構來闡述屬性之間的關聯及強弱程度，最後透過切割圖形達到分群的效果。本作法所呈現的族群屬性與量測結果，不同於一般以核心距離偵測或向量比對的做法。我們獲得的群聚反應具有精確、易於理解及快速執行等優勢。 With the electrifying of the Internet, all types of electronic information have been rapidly growing and increasing. Although massive amount of research has been dedicated to information searching, it seems that we are faced with the awkward situation of barely permissible results and deadlocked progress, and we are hoping to advance search utilities with higher accuracy, better practicality, and greater convenience. In multitudinous research topics, clustering is often used to classify large-scale information, and has had exceptional results, but clustering is rarely used in search engines; therefore, this paper will discuss utilizing cluster in search technology. We apply the Association Rule to pull closer the relation between common attributes, and then use a Graph Model structure to elaborate on the association and strength of each attribute, and lastly we adopt graph segmentation to achieve classification. The category attributes and test results displayed by this approach are different from those achieved via normal distance detection, and the obtained results have advantages such as higher precision, easier understandability, and faster execution.
