Enterprises have collected large bodies of data that describe interactions between their customers. Consider the graph of telephone calls and SMS for a Telco, connections within emails and P2P services for an internet service provider, and links between payments by customers of financial institutions.

Relational databases are fundamentally unfit to explore the graph within a social network and Big Data solutions (Hadoop, etc) are usually not meant to work with sparse graphs. The maturing capabilities of Graph Databases have made them the optimal approach to mine these social networks.

During this presentation, we will discuss applications of graph mining using two datasets.

One data set contains 2 million telephone users over a period of 30 days. The data consists of telephone calls and SMS. In addition we have for each person their physical location for most of the day. We will show how we analyze the social, geospatial and temporal information to create deep insights into the customer’s behavior.

The second dataset is anonymized information from an on-line bank in Asia. The data includes all payments from account to account along with details about links to each other through IP addresses, goods traded, location, etc. We will show how we can detect, in real time, whether an account executing a transaction is part of a group of accounts that is somehow linked to fraudulent activity.