The Detection Of Artificial Intelligence Frontier

The accurate judgment in the frontier fields of science has a bearing on the future of science,technology and innovation in a country.In recent years,the research of artificial intelligence,as as a major national development strategy of the world's major scientific and technological power including China,has been drawn worldwide attention.Scientific papers and patent documents,as forms of basic scientific research and corresponding technological innovation,provide us with rich data bases which can be used in many scientific research innovation activities.Based on the scientific papers and patent documents in the field of artificial intelligence,the systematic analysis indicators and exploration model of research frontiers identification are established to identify,select and interpret the research frontiers in the field of artificial intelligence,facing different detection needs of research frontiers.Meanwhile,the typical cases of deep learning and other research frontiers are selected for detailedly analysising.Firstly,this paper systematically investigates the theories,methods and practical applications of research frontier of domestic and foreign research,defines the research frontier based on summarizing the predecessors' concepts of research frontier,and analyzes the frontier detection methods of scientometrics and computer mining.According to the current exploration needs of hot,emerging and breakthrough research frontiers,the systematic analysis indicators in four dimensions of novetly,innovation,interdisciplinarity and high degree of concern and exploration model of research frontiers identification are constructed,and a set of comprehensive methods to automatically mining the research frontier of science is put forword.Secondly,based on the constructed research frontiers identification systematic analysis indicators and exploration model,this thesis takes the top 1% of scientific papers cited in the field of artificial intelligence from 2006 to 2016 as the research object,using hot word analysis,quotation of burst terms monitoring and co-occurrence analysis to mining and identify the research fronts.This paper calculates the novelty index based on the literature year of publication,innovative index determined by strength of burst term detection,interdisciplinary index according to the literature classification,and high degree of concern index reflected by total citations and the average cited frequency per paper.The thresholds are set according to the calculation results to select hot front research frontiers,emerging research frontiers,and breakthrough research frontiers.In interpretation stage of research frontiers,the relevance of science and technology and media attention are introduced for a detailed explanation from different perspectives of research frontier fields,detailly interpreting the research frontier cases form the topics,core documents,indicator characteristics,detailed description,country distribution,distribution of research institutes and core researchers.Finally,typical cases are selected to verify and demonstrate the latest advances in the field of artificial intelligence.The research results of this paper show that the topics of key hot research frontier include support vector machine algorithm,its improved algorithms and applications for solving classification and multi-classification problems,and intelligent optimization algorithms for multi-objective optimization and global optimization.The topics of key emerging research frontiers include deep learning networks and time-varying delay neural networks and its stability and synchronization research;The topics of key breakthrough research frontiers include deep learning networks,group decision-making and particle swarm optimization algorithms and its improved algorithms.The degree of scientific linkage in deep learning area is 1.82,which is higher than the same period of scientific relevance of graphene.The time lag of patent citations in deep learning area is 7.4 years.From the perspective of social media,solving the problems of asymmetric information and system autonomous learning in real life are drawn more attention.