Establishing an Emerging Topic Multi-Indicator Detection Method: A Case of Deep Learning Field

• Main Authors: LAI, JIA-JIE, 賴加傑
• Other Authors: CHEN, SSU-HAN
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/77dw66

碩士 === 明志科技大學 === 工業工程與管理系碩士班 === 107 === Detecting emerging topics has become a hot research direction in recent years. Artificial Intelligence (AI) is the hottest topic of discussion in technology and Deep Learning is an indispensable technology in AI. This research queries the term “Deep Learning” in Web of Science (WOS) database and relevant conference papers and their bibliographic are retrieved. A series of research methodologies include bibliometric, social network analysis, clustering and multi-indicators are designed and implemented via bibliometrix package in the R environment to achieve the goals of this study. Each topic is evaluated the degree of emerging by six indicators: citing half-life (CHI), endogeneity index (EI), originality index (OI), multidisciplinarity index (MI), co-authorship (CA), and public sector participation (PSP). This study will build a network at each time point, cluster the network, calculate multiple indicators for each cluster, and then integrate multiple indicators into the emerging degree and form a Rice distribution. In the future, emerging topic is identified by observing whether the degree of emerging of an oncoming cluster is significant or not. This method provides a reference of what the future trends will be which is able to guide scholars, decision makers, or investors to understand the direction of development.