Research on Emerging Topic Recognition and Feature Association Based on Topic Model and Time Series Analysis

Abstract

[Purpose/Significance] Carrying out research on emerging research topics(ERT) identification and scientifically and effectively discovering their characteristic correlation laws can better serve practical needs and give play to the innovative supporting role of sci-tech information research on the development of disciplines. Aiming at discovering emerging research topic(ERT) and its characteristic correlation effect scientifically and effectively, this paper carries out ERT identification and feature analysis, while realizing the innovative supporting role of sci-tech information work. [Method/Process] Starting from the definition of the features of ERT, this paper established the methodological framework of ERT identification by using natural language processing, global principal component analysis and time series analysis. Based on the relevant theories and practices of emerging topic identification and scientific impact assessment, this thesis quantified the characteristics of the topic’s consistency, novelty, influence, and growth. On the basis of emerging themes identification, the law of the development of emerging themes in the target field is deeply excavated. Granger causality test and cointegration analysis were used to explore the long term equilibrium and the correlation effects of their characteristics. [Result/Conclusion] This paper proposes a method to identify ERT and their correlation feature analysis. In order to verify the effectiveness and feasibility of this method, the field of wetland was selected to carry out empirical research. Combined with the topic identification and feature correlation effect analysis, the final result depicted the dynamic development path of subject science influence in this field, while putting forward some advices on developing emerging topics from the perspective of associated characteristics.

Keywords

• trend forecasting