Location based social networks

Efficient embedding for hypergraphs

My work was centered around the application of hypergraphs to location-based social networks (LBSNs). Under the guidance of Prof. Srijith P K and Manisha Dubey, I started with implementing the LBSN2Vec algorithm in Python. For LBSN, the network can be modeled into a k-uniform hypergraph, so I compared the results of location prediction using a perceptron based architecture, DHNE and a random walk based architecture, LBSN2Vec. The LBSN2Vec algorithm uses the social network as an additional information as compared to the DHNE algorithm. In order to remove that, I modified the former to not use the friendship graph. Instead, to achieve the random walk with stay procedure, I operated the random walk on a shuffled sequence of locations ids and for each node on that walk, sampled nodes from the other 3 domains.