|
本研究提出一種透過專利引證網路及主成分分析來補強專利技術/功效矩陣圖之方法，藉以使技術/功效矩陣中的檢索廣度增加而不影響其原本的準確率，此方法主要可分為三階段。第一階段係運用專利引證網路分析美國專利資料庫(USPTO)中之特定檢索標的，並以共引證分析之概念找出一群與檢索標的有關聯之候選專利群；於第二階段中，本研究以國際專利分類號做為技術判斷依據，加以計算出各候選專利與各檢索標的專利之相似度，並建立一相似矩陣；第三階段係針對該相似矩陣進行資料分析，本研究以主成分分析法萃取其特徵值和特徵向量，並透過主成分之係數找出與檢索標的最相似之專利。最後，本研究提供兩個實際案例做為驗證，亦證實利用此方法可獲得與檢索標的高度相似之專利群。
關鍵字：技術/功效矩陣、專利引證網路、專利相似度、主成分分析
The technology-function matrix
is the main tool for analyzing patent portfolios in aspecific
technology field; however, the quality of a technology-function
matrix is tied toinitial patent search. This thesis provides a
method of reinforcing thetechnology-function matrix of patent
map, so as to improve the retrieval recall ratiowithout
affecting the precision ratio thereof. More specifically, the
method comprisesthree phases: Phase I: generating a group of
candidate patents by applying patentcitation network and
co-citation analysis within USPTO database; Phase II: creating
asimilarity matrix by evaluating patent similarity between the
group of original patentsand candidate patents; and Phase III:
Obtaining a reference list of patents that are highlysimilar to
the group of original patents by principal component analysis.
Furthermore,the thesis takes Through-Silicon Vias (TSVs) and
Machine Tool Diagnostic System astwo case studies to verify the
feasibility and effectiveness of the proposed method inthis
thesis.
Key words: Technology-function
matrix, Patent citation network, Patent similarity,Principal
Component Analysis (PCA)
|
