台灣科技合作研究社群複雜網絡結構特性分析

Abstract

台灣科技創新體系內為因應現階段科技研究創新政策進而持續投入大量研究經費，但是資源有限條件下之科研經費應如何有效且合理地分配給進行科研計畫的廣大社群成員卻是個難題，本研究藉由台灣國科會GRB搜尋系統各年度合作研究複雜網絡之資料集轉換成為有效網絡數據，運用UCINET網絡分析平台進行內聚力、小團體、自我網絡、中心性、角色位置與冪次定率等分析程序以及NetDraw探勘繪製出各年度複雜網絡結構圖形並以SPSS對各年度樣本作描述統計分析，實證分析各年度網絡內個別大小社群間連結特性，得以解決目前國科會科技計畫績效管考與GRB搜尋系統等平台難以單獨地將各年度合作研究複雜網絡社群結構作驗證及辨別區分與如何將其重疊結構分形性圖形化出來等問題。 就GRB搜尋系統歷年來資料集內六大研究領域暨八十一個次研究領域，彙整為實質網絡數據，依據本研究所提出「以計畫主持人（PI）為網絡結構重疊模體（motif）三維模型」為電腦模擬分析模型基礎，進而形成各依年度區別之台灣科技合作研究社群複雜網絡整體結構，探勘完整科研合作研究社群網絡重疊結構暨個別自我網絡模體，分析相關複雜交互影響作用的結構特性並加以圖形化。 本研究完成台灣科技研究社群之複雜網絡整體結構與個別重疊社群結構之特性驗證分析，驗證出「無尺度」、「小世界」與「分形性」等複雜網絡特徵，發現並建議以「集結節點」、「群體中心性」及「門檻比率」作為調控管理下一年度合作研究社群網絡結構之重要參考依據。

At present, the essence situation of Joint Research Projects Community in Taiwan lacks a complex networks structure characteristic analysis as definite as in Scale free, Small world, Fractal, et cetera. In the Joint Research Projects Community networks, there are many subgraphs and overlapping structures inside them. The thesis research designs the multiple dimensions ego network motif of principal investigators and readies for exploring all motifs of annual real networks data sourced by the government research bulletin of Taiwan National Science Council. The thesis research in estimating, analyzing and graph-drawing for the Joint Research Projects Community networks in Taiwan, it makes use of SPSS l5.0, UCINET 6.186 and NetDraw 2.076 for Windows to compute all real data of sampling survey, after analyzing to demonstrate the complex networks structure characteristics with several analysis processes such as Cohesion, Subgroups/Cliques, Centrality Measures, Group Centrality, Roles and Positions, Power Law and Networks Graph. The conclusion reveals a special significance of specific projects management, means that the group centrality and threshold rate of each annual Joint Research Projects Community of Taiwan, should be monitored and operated it in coordination with specific projects member and drawn up new policies of the authorities, then the national innovation system resource will be effective to distribute in the future. The thesis research successfully tests and verifies all complex networks structure characteristics of Joint Research Projects Community, also draws out all subgraphs and the entirety graphs in each year.