Reducing Computational Complexity of Network Analysis using Graph Compression Method for Brand Awareness Effort

Abstract

Online social media provides platform for social interactions. Thisplatform produce large-scale data generated mostly from onlineconversations. Network analysis can help us to mine knowledge andpattern from the relationship between actors inside the network. Thisapproach has been crucial in supporting prediction and decision-makingprocess. In marketing context such as branding effort, using large-scaleconversation data is cheaper, faster and reliable comparing mainstreamapproaches such as questionnaire and sampling. Social network analysisprovides several metrics, which was built with no scalability in minds,thus it is computationally exhaustive. Some metrics such as centralityand community detections has exponential time and space complexity. Withthe availability of cheap but large-scale data, our challenge is how tomeasure social interactions based on those large-scale data. In thispaper, we present our approach to reduce the computational complexity ofsocial network analysis metrics based on graph compression method tosolve real world brand awareness effort.