A new automatic approach for research trend analysis based on scientific text mining

Research trend analysis for a specific research area (through different time frames) can lead to a better understanding for researchers in that area, and for policy makers in research for contributing in assigning research funds and policies. An important and practical approach to analyzing research trends is to study and review research data and publications using scientometrics and research document processing. In this research, we have proposed a text mining approach for analysing research publications in a specific area, in order to analyze and identify important research topics. In this paper, we propose a method to analysis a scientific trend. The proposed approach is based on clustering keywords through a new co-occurrence matrix. A new metric is also adopted to identify centrality and density (maturity) of a specific research area and identify keywords and contributory topics. For achieving this, we proposed a trend analysis method using co-word matrix with and for deeper analysis we use clustering and strategic diagram with propose indices. In order to test and evaluate the proposed method, we use comparative evaluation method. In addition, for more analysis, we have selected research publications in a specific period of time (2012-2016) in Mechanical Engineering, which are extracted from WoS (Web of Sciences) database. We have applied the proposed metrics to evaluate research trends and identify contributory areas and topics in the selected documents. The comparative evaluation shows an improvement in proposed trend analysis method.