In this project we attempt to build a graph based video recommendation system without using graph neural networks. We experiment with three methods of link prediction: 1. using a variant of random walk, 2. generate node embeddings using random walks for link prediction 3. employ a topology-centric framework (GELATO) to achieve desirable results for link prediction. We also analyse the scalability of our approach and try to make existing work scalable to larger graphs. We present a comparison between using non-cluster approach and clustering the graph and running the same methodology for the methods mentioned previously.
For more details about the results and the methodology refer to the attached report.