Enhanced link prediction using sentiment attribute and community detection

In social network analysis, link prediction is an important area where the researchers can find the missing links and the future links possible among the users. Often, link prediction is made by analyzing the social linkage of the users in the given networks, i.e., the Topological structure of the networks. However, this approach leads to inconsistencies when researchers want to emphasize topics on which users have mainly engaged their selves in discussions. Mainly, this approach predicts future links based on available network structures without considering the topics on which the users are participating. This can be enhanced by incorporating the sentiment attributes and the community structure of the users in the network. In this paper, we propose an algorithm that incorporates the sentiment attribute of users and community structures along with the topological features. To evaluate the same, we have crawled the tweets of various countries concerning COVID-19 from Twitter. Experimental results show that users exhibiting the same emotion and belonging to the same community will influence other users to connect, thereby improving the performance of the link prediction.