「自動化與控制系統」領域中人工智慧文獻之資訊計量與研究前沿分析

1956年由John McCarthy正式提出人工智慧一詞，確立此一新興學科後，相關研究即迅速成長。人工智慧為一跨領域之學科，人工智慧的發展也為諸多領域帶來突破性的成長，自動化與控制系統領域可謂首當其衝，甚至可說人工智慧技術是自動控制系統運行的基礎，人工智慧進入自動化與控制系統領域後，產生了智慧控制，改善工業自動控制系統的問題、提高自動化的效率，也促成了工業4.0時代。儘管目前已有不少人工智慧文獻的資訊計量學研究，但尚未有人工智慧應用於「自動化與控制系統」領域的相關資訊計量學分析，故本研究自SCIE資料庫檢索1956至2019年間「自動化與控制系統」領域中的人工智慧文獻並下載書目資料進行資訊計量學分析，以綜觀其文獻特性與研究前沿。本研究歸納1956至2019年間「自動化與控制系統」領域中的人工智慧文獻分析結果如下：(1)整體文獻呈指數成長模式，但不同國家呈現不同的文獻成長模式；(2)交叉研究領域以電機電子工程為主；(3)產出最多篇文獻的作者為Ge, Shuzhi Sam，多數高生產力作者以平等法計算出的文獻產出數量遠大於以第一作者法計算出的文獻數量；(4)產出最多篇文獻的機構為中國科學院，且中國有最多文獻產出排序在前100所的機構，中國的文獻產出篇數也為所有國家中最多，文獻高度集中在少數國家；(5)刊載最多相關文獻的期刊為《Robotics and Autonomous Systems》，由IEEE與Elsevier所出版之期刊刊載多數文獻；(6)多數文獻未曾被引用，文獻被引用次數最多的作者為Pedregosa, Fabian，被引用次數最多的機構為美國的加州大學系統，美國與中國被引用次數排序在前100所的機構數量最多，被引用次數最多的國家為美國，被引用次數最多的期刊為《IEEE Transactions on Automation Science and Engineering》；(7)被引用次數排序會受到文獻產出數量、平均每篇文獻被引用次數、是否產出高被引文獻與未被引用文獻數量等數據而有所差異；(8)期刊分布符合布萊德福定律，且與布萊德福─齊夫定律曲線圖吻合，進一步驗證核心期刊後發現，布萊德福分區所區分出的三區期刊，僅能彰顯出第一區影響力優於第三區，但大致可辨識出核心期刊；(9)作者生產力不符合洛卡定律、普萊斯平方根定律，此外作者與期刊生產力不符合80/20定律，機構與國家生產力則符合80/20定律；(10)作者關鍵詞之共現分析以「神經網路」與「自適應控制」為重點作者關鍵詞，但「神經網路」的關注度已逐漸被後者取代；(11)文獻共被引以Slotine, J.的著作《Applied Nonlinear Control》總連結強度最強，作者共被引以Ge, Shuzhi Sam為主的叢集內彼此連結強度較強。期刊共被引以《Automatica》的總連結強度最強，《IEEE Transactions on Automatic Control》次之，且兩者連結強度強；(12)作者合著研究中有二位以上作者合著一篇文獻的比例達91.323%，Ge, Shuzhi Sam叢集內的作者合著關係較強；機構合著研究中有二所以上跨機構合著的文獻為46.858%，亞洲國家機構合著關係較強，相較於亞洲國家機構進入此研究領域較晚，歐美等國在此領域發展較為成熟。國家合著研究中有二個以上跨國家合著的文獻為21.507%，中美兩國合著關係最強。本研究依據前述研究發現，針對資料庫、視覺化軟體、圖書資訊服務界、「自動化與控制系統」領域人工智慧研究人員，以及相關產業人士與政府提出相關建議，最後提出後續研究建議包括擴大資料庫檢索範圍與資訊計量學範疇研究、以不同計算方式綜整比較生產力與被引用次數、比較多種視覺化軟體並驗證分析結果之可信度、針對領域學者與產業界人士進行訪談與質性分析。

關鍵字

人工智慧；自動化與控制系統；資訊計量學；研究前沿；知識圖譜

並列摘要

“Artificial Intelligence” was formally proposed by John McCarthy in 1956, and related research grew rapidly from then. Artificial intelligence is a cross-disciplinary topic, and the development of artificial intelligence research has also brought breakthrough growth in many fields. “Automation and Control Systems” was the field directly encounter drastic impact from AI, and it can even been said that artificial intelligence technique is the foundation of the automatic control system. AI enters this field and bring in the intelligent control, improving the problems of industrial automation and control systems, increasing the efficiency of automation, even contributed to the coming of the Industry 4.0 era. Although there have been many informetrics researches on artificial intelligence literature, there is no relevant informetrics analysis on artificial intelligence applied to the field of “Automation and Control Systems”. Therefore, this study analyzes the characteristics and research fronts by informetrics approach of artificial intelligence studies articles in the field of “Automation and Control Systems” during 1956-2019 retrieved from SCIE database. The results of this study show that artificial intelligence studies articles in the field of “Automation and Control Systems” during 1956-2019 have some findings as follow. (1) The growth pattern of the articles fits exponential growth, but differs from country to country. (2) The studies are including multiple disciplines, and most of Electrical and Electronics Engineering. (3) The most prolific author is Ge, Shuzhi Sam, it’s regular in most highly productive authors that the number of articles counting every author is much more than only counting the first author. (4) The most prolific institute is Chinese Academy of Sciences, China has the most number of institutes rank among top 100 sorted by the number of published articles, and China is the most productive country. The articles are highly concentrated in a few countries. (5) Robotics and Autonomous is the journal which has published the most number of articles, most of the articles published in the journals published by IEEE or Elsevier (6) Most articles haven’t been cited. The most highly cited author is Pedregosa, Fabian; the most highly cited institute is University of California System in the USA, the USA and China have the most number of institutes rank among top 100 sorted by times cited counts, and the most highly cited country is the USA; the most highly cited journal is IEEE Transactions on Automation Science and Engineering. (7) The rank sorted by times cited counts is affected by the number of published articles, average citations per article, the number of highly cited articles, and the number of uncited articles. (8) The distribution of journals fits Bradford Law and Bradford-Zipf’s curve. The three regions of journals distinguished by Bradford Law has significant difference only between region one and three, but it can still identify the core journals. (9) The distribution of authors neither fits Lotka’s Law, nor Price Square Root Law; the distribution of authors and journals don’t fit 80/20 Rule, but the distribution of institutes and countries fit 80/20 Rule. (10) In the analysis of author keywords co-occurrence, “neural network” and “adaptive control” are the main author keywords, but the focus of the former is gradually replaced by the latter. (11) Applied Nonlinear Control wrote by Slotine, J. has the most total link strength counts in co-cited references analysis; the authors in the cluster represented by Ge, Shuzhi Sam have stronger links between each other in co-cited authors analysis; Automatica has the most total link strength counts in co-cited journals analysis, and IEEE Transactions on Automatic Control is second, there is a strong link between them. (12) In co-authorship analysis, there are 91.23% articles have more than two authors, and the authors in the cluster represented by Ge, Shuzhi Sam have stronger links between each other; there are 46.86% articles have more than two institutes participated in, and the institutes in Asian countries have stronger links between each other. Compare with Asian countries which entered this research field late, the Occident countries are more well-developed in this field; there are 21.51% articles have more than two countries participated in, and there is a strongest link between China and the USA, China has also strong link with Singapore and Australia. This research makes a suggestion to databases, visualization software, libraries and information services, artificial intelligence researchers in the field of "Automation and Control Systems", relevant industries and governments based on aforementioned research findings. At last, this research proposes some recommendations for follow-up researches including expanding the retrieved scope of database and other part of informetrics analysis, comprehensively comparing productivity and citations with different calculation methods, comparing different network visualization softwares and verify their effectiveness, interviewing artificial intelligence researchers and experts in the field of “Automation and Control Systems” and carry on qualitative research.