調整與運用h指數於研究與專利績效評估之探討

Abstract

h指數自2005年提出後形成一個相當擁擠的研究領域，並已獲Web of Science、Scopus等商業資料庫所採用，h指數因此成為幾如標準的研究績效評量指標。本論文的研究係從對h指數的一種幾何的闡釋上出發。基於此一幾何闡釋，我們設計了二個擷取形狀特徵的c-descriptor與 t-descriptor來分別代表h-core area與h-tail area的特徵高度。經由實證資料驗證，h指數如果輔以c-descriptor與t-descriptor，對於研發與專利的績效評估具有相當好的區別能力。在此成功基礎上，我們進一步利用h-core area與h-tail area的形心來分別代表h-core與h-tail的績效，使得大量被評估對象的研發與專利績效，可以同時由其影響面與產能面觀察與比較。在此幾何基礎上，我們開發了二種靜態分析的作法：(1)可以快速定位特定被評估對象在大量的對象中的位置、以及(2)可以將大量的被評估對象快速的予以定性分類。接著我們繼續將形心的概念繼續拓展納入沒有引用數的文件、以及觀察形心在一段時間的位置變化所反映出的績效演進軌跡。同樣基於相同的幾何闡釋，我們透過視角的轉換，認為一個領域的h指數，可以視為是該領域的最顯著的文件的集合所導致的。這個最顯著的文件的集合則可以視為是所有相關被評估對象所共同貢獻的，因而我們即以被評估對象對此最顯著的文件的集合的貢獻比例作為其在該領域的績效。我們因此提出了一個概念架構以及二個主要實施這個架構的方法。透過實證資料，這個架構具有合理、直覺易懂、對個別被評估對象可以獨立解釋、而且可以一體適用於特定領域與跨領域的績效評估。

In this Dissertation, we start with a novel geometric interpretation to the ranking by h-index. Based on this interpretation, we design two shape descriptors, c-descriptor and t-descriptor, to represent a characteristic height for the h-core and the h-tail areas, respectively. With the verification of empirical data, we have demonstrated the differentiation capability of the h-index supplemented with the c- and t-descriptors to both innovation and research performance evaluation. We have further designed a two-dimensional scheme using the centroids of the h-core and h-tail areas with which a large number of individuals’ research or innovation performance can be observed and compared from their impact side and productivity side simultaneously. We have also developed two static analysis approaches for (1) quickly positioning a target individual’s performance; and (2) conveniently partitioning a group of individuals into qualitative categories. Subsequently, we further extend the area centroid idea to incorporate the un-cited articles and to trace the trajectories of area centroids over a period of time so as to monitor the performance evolution. Also based on the geometric interpretation, we have started a new line of research concerning the research and innovation performance evaluation spanning a number of fields. By twisting the perspective to the geometric interpretation, we consider that it is a set of the most visible articles of a field to result in the field’s h-index. We then consider this visibility set as a joint contribution from the accessed individuals, and determine their performance in the field as their ratios of contribution. A conceptual framework and two major implementations are thereby designed. We have found that the conceptual framework is reasonable, intuitive, self-contained, and applicable to both field-specific and cross-field performance evaluation in a unified manner.