­

Application

A software model for generating colocalization information from mobile data and constructing a social network.

Key Benefits

• Software that can be licensed.
• Constructs social networks from any mobile location data.
• Largely reducing the total computational time by over 20 times.
• Scalable to extremely high data volumes.

Market Summary

The explosive proliferation of detailed location data from mobile devices has enriched marketers’ potential to construct consumer profiles with data on their physical movements. While marketers’ knowledge of consumers’ social ties has been fueled by online social networks, which have been used as a means of spreading marketing messages, large-scale location data gathered from mobile devices now offer a similar opportunity to examine offline networks. Large multinational companies as well as leading hospitals have collaborated with advanced analytics and machine-learning solution providers to deploy solutions that can improve their predictive analytics, virtual assistance and, respectively, consumer and patient-monitoring solutions.

Technical Summary

Emory researchers and colleagues have created an algorithm designed to efficiently process large-scale mobile location data for co-location metrics and construct a social network which can be useful targeted marketing or for identifying influencers – specifically offline influencers. This analytical tool can quickly observe, identify, and construct meaningful social networks from offline mobile data, and can be used for marketing or research. This network construction procedure is optimized by using an elastic search algorithm with multi-thread parallelization, conducted on multiple virtual machines simultaneously. The optimization reduces the computation time for the “simple” approach from more than 50 hours to less than 2 hours. This can serve as the foundation for any work that looks at social influence, as well as predictive models that seek to leverage social connections.

Developmental Stage

The research team is currently testing the incorporation of the co-location network into predictive models of the locations visited by individuals. The investigation aims to identify the incremental value of the co-location information and demonstrate its applicability for targeted marketing efforts. Pending this work, the research team intends to incorporate the co-location network into estimation of individuals’ social influence, which can aid in the identification of offline influencers with which marketers may partner.