Predicting Research Trend Based on Bibliometric Analysis and Paper Ranking Algorithm

One of the most essential investigation demands in the computational evaluation of scientific publications is whether the immense collections of scientific papers hold significant indication about the dynamics included in the development of science; signs that may help predict the growth and decline of scientific methods, ideas, and even fields. The research presented in this paper focuses on a general approach to analyze and predict the thematic evolution of a given research field by pointing out uptrend keywords. In particular, we propose a dataflow of method and a paper ranking algorithm. This results in ranking papers, from there we select the best 20 papers and extract from them meaningful keywords (concerned with research field/subfield, algorithms, methods, etc.). After that, we formulate the final score for keywords by summing up scores of papers containing them and then group the obtained results by year. Therefore, we can demonstrate scores of keywords through years in time series and observe which keywords are displaying an upward tendency. As a case study, the proposed approach is applied to analyze the thematic evolution of the Artificial Intelligence research field in the period 2005–2016 from the Web of Science database. Ultimately the method is evaluated by checking occurrences of predicted keywords in true prominent keywords in timeframe 2017–2019 and provides precision 73.33%.