Predicting Research Trends by Using Link Prediction in Keywords Network

Please enable javascript in your browser.

Behrouzi, Saman | 2017

542 Viewed

Type of Document: M.Sc. Thesis
Language: English
Document No: 49904 (52)
University: Sharif University of Technology, International Campus, Kish Island
Department: Science and Engineering
Advisor(s): Hajsadeghi, Khosrow; Kavousi, Kaveh
Abstract:
The rapid development of scientific areas in this modern era makes the process of finding new field of research slow and laborious for prospective scholars. Thus, having a vision of the future could be helpful to pick a right path for doing researches and ensuring that it is worth to invest in. This thesis seeks to predict research trends by using link prediction approaches on keywords network and discusses about the performance of various algorithms in different situations. Moreover, for the last part of the experiments, novel link prediction algorithms are proposed by the author, enhances the accuracy of prediction results. The data set collected from Sciencedirect and Scopus by a strong web scraper from more than 200 computer science journals, contains more than 345,000 nodes (author defined keywords) and roughly 1.8 million links until year 2016. Experiments on large keywords network suggest that information about future interactions between keywords can be extracted from keywords network topology alone, and the proposed metrics for link prediction have more accurate prediction than the former metrics
Keywords:
Link Prediction ; Keyword Extaction ; Keyword Network ; Research Trends Prediction

No TOC