近幾年來,運用專利分析法於預測新興技術這方面已經有許多相關的應用與研究問世。尤其是在專利分析的領域之中,引證法扮演了一個非常重要的角色。然而,過去引證法在專利分析上仍有些缺點,像是無法考慮到整個專利之間相互引證的關係,並且在處理上是耗費時間的。因此,本論文便利用專利引用網路圖中來改善專利之間相互引證的關係,並透過關聯矩陣來分析專利引證中的潛在有用資訊。此外,為了更有效率地辨識出具關鍵影響力的專利文件,以及改善專利引證的關聯分析,本文將試圖透過結合topological sorting演算法、triangularization演算法和網路測量法這三種方法於關聯矩陣分析上來做分析的動作,並且試圖最佳化。最後,並透過探討本文所提出的專利分析模組的結果,依其模式的判別提供更有效率的建議與分析。
In recent years, there are many applications for patent analysis to forecast emerging technologies. Especially, citation plays an important role in the patent analysis field. However, current patent citations have certain drawbacks, such as that is difficult to grasp the overall relationship among patent documents, and citation analysis is a time-consuming task because it requires an exhaustive search. Therefore, this thesis applies an incidence matrix to analyze the citation information in patent citation networks. In order to analyze the patent citation and identify the critical patents more effectively, the matrix is analyzed using two algorithms, the topological sorting algorithm and the triangularization algorithm are combined with network measurements as a result of those contributions. Finally, some general discussions of patterns from the results of a patent citation model are presented along with conclusions.