Unsupervised machine learning method for researchers’ profiles matching

Abstract EN

Researcher profiles matching is an initial and important step of effective research teams’ formation.

the researchers’ wide, multidisciplinary, and changeable research interests complicate the process of profile matching using traditional methods and affect its performance.

this research aims to solve the problem of profile matching in scientific research and scholarly work by employing unsupervised machine learning methods.

the k-mean clustering method is utilized to categorize researcher profiles based on the statistical analysis of their publication titles, and the correlation-based similarity is employed for profile matching within the categories.

the proposed method is implemented, tested, and evaluated using an extracted dataset from google scholar.

the profile matching results and the clustering quality test result show that the designed task was achieved, in addition to high similarity values of publications within the categories and low correlation values among the clusters.

moreover, the clustering results’ analysis can reveal interesting and enlightening information about the scholarly work, which may help the researchers, research management departments, as well as policies and decisionmakers in their scholarly work associated tasks.