Modelling Weighted Social Network Graphs from User Interactions

Link to slides - https://docs.google.com/presentation/d/1v1DsTSBPLRNsh_0DOvNyfOoMQJk1TZmdEFaviOUJr8k/edit?usp=sharing

Social network analysis traditionally makes use of unweighted graphs, with no other information between peers except for acknowledgement of existence of a link. In our work, we attempt to construct a weighted graph using user interactions, where weights represent the degree of recent frequent interaction. We reward the links among users for each interaction and also decay those rewards over time. We therefore, use such recent interactions as evidence of a stronger bond and assign weights accordingly. Finally, we optimize our values of rewards and decay using some power laws which are exhibited by real world weighted networks