Hybrid DBSCAN based Community Detection for Edge Caching in Social Media Applications

Social media applications offer multimedia content to enrich the user experience in pocket-sized mobile phones. The highly diverse features of social media applications put pressure on the battery life of smartphones. Moreover, accessing social content from cloud-based infrastructure involves greater access latency because of the high geographical distance between the cloud and mobile users. To meet the increasing demands of lower energy and latency applications, caching in mobile edge computing plays a vital role by pushing processing and storage resources at the edge of the network. Mobile edge computing offers minimum delays as it provides data content near to the end-user in a ubiquitous environment. However, not all the content can be cached at the edge node. We propose a community-based clustering framework that is used to identify the users having similar interests. For community detection, we propose a hybrid framework with a combination of minibatch K-means and DBSCAN. The framework scales well in terms of the size of the data set. We determine a set of popular social content from each community and cache them at the edge of the network. In comparison to traditional cloud and popularity based content delivery schemes, our proposed edge-based framework provides lower access latency and higher smartphone battery lifetime.