Co-Author Link Prediction from Academic Databases

Abstract

Social networks: social communication platforms where individuals exhibit their attitudes and behaviors, gestures and mimics in symbolic actions in a virtual environment. Their swift development and wide usage proportion rises social networks' necessity and popularity. People communicate with each other and share some data like idea, thinking, photo, video and location. When this data is processed, very important information is obtained about the users. In this light, it is possible to make important estimates about the users. To analyze a network, the network model, properties of graph structure, criteria and metrics must be calculated first. In this study, similarity calculation between the subjects which authors studied has been done. In this application, five suitable models (Jaccard Index, Sorensen Index, Common Neighbor, L. H. Newman Index and Salton Index methods) are applied to the data set generated by randomly selected twenty writers. In order to evaluate the calculation results; the results of the similarity calculation between authors with known linkages are used as reference values. Comparisons of obtained similarity values with reference values result in new connections. These links reveal that authors can publish articles together.