The mobile phones that we use everyday and carry around everywhere with us, create huge amounts of data that trace the daily patterns of our behavior. The interactions we have with others through calls or messages map out our social relationships, business dealings and interactions with the wider community as complex interconnected circles of calls.
This data visualization is created using mobile phone subscriber calling patterns. Each dot (or node) represents a phone number that is called by a subscriber, the larger the node size the more often it is called. The lines (or edges) between nodes represent a call from one number to another.
Each subscriber will have a unique calling pattern that can be used to develop pricing plans, identify him or her and can even predict his or her behavior. For instance a subscriber that is in the process of switching to a different network provider will show up as two similar patterns one from an on-net number and one from an off-net number.
This particular chart was produced at the early stage in a series of analytics and was used to filter out the first level of calling patterns types. The data used here represents a very short period of time, just a few seconds. We can see at the top right-hand side of the graph large loops that show numbers, which have been called many times in this short period. These are likely to be machines, such as the auto dialer systems that use pre recorded messages when answered, Interactive Voice Response (IVR) systems, security systems and alarms. Humans would not be able to make so many calls so quickly. These numbers were isolated out as a separate segment and subsequent analysis focused in on the detailed individual human calling patterns.
This visualization shows a representation of a graph, although the layout parameters have been used to create a format that is unlike those typical used to display graphs. An issue commonly faced in this area is that the connected graphs quickly become huge and are almost impossible to visualize due the sheer number of callers and interactions. To take a sample from a highly connected graph is a difficult problem, as we need to decide which connections to ignore. In this case a very short period of time is used to cut down the output to a manageable size.
The underlying data format is rather simple, calling number, called number, time of day and duration. The data is first clustered using a machine-learning algorithm to create the groups and then displayed as a graph using Aster Lens.
Christopher Hillman is based in London UK with his wife and two kids and is a Principal Data Scientist in the Advanced Analytics team at Teradata travelling extensively in the International Region.
His passion for analytics spans 20 years of experience working in the business intelligence and advanced analytics industries. Prior to Teradata Chris specialized in the Retail and CPGN vertical, working as Solution Architect, Principal Consultant and Technology Director. Chris currently works together with the Teradata Aster Centre of Expertise and is involved in start-up analytics for Big Data projects helping customers to unlock insights from their data including understanding where MapReduce or SQL is an appropriate technique to use.
As well as working for Teradata, Christopher is currently studying part-time for a PhD in Data Science at the University of Dundee applying Big Data analytics to the data produced from experimentation into the Human Proteome. His research area involves real-time analysis of Mass Spectrometer data using Parallel algorithms. Part of his duties at the University include lecturing on Hadoop and MapReduce coding.
You must be a registered user to add a comment. If you've already registered, sign in. Otherwise, register and sign in.