Social Network Analysis
Note: Code and details not shared due to non-disclosure agreement
Introduction
Social Network Analysis is a form of analytics which utilizes network and graph theory to uncover deeply hidden patterns which exists between numerous data points. The most famous algorithm which uses Social Network Analysis is the Page Rank Algorithm which was created by the founders of Google. Social Network Analysis can be used to undercover relationship between any type of social networks data (Facebook,LinkedIn, Twitter) or call data records (incoming, outgoing calls). When we dive deeper we discover not just the relationship between two points, but how they relate to superstructures within the data.
In order to utilize these metrics we need to know the following:
-
Node (The actor) - Incoming Call
-
Connection (The edges) - Outgoing Call
The common features which are utilized for analysis include but are not limited to the following:
- Degree Centrality: Is the score assigned to the number of links between each node
- Betweeness Centrality: Is the score assigned to the number of times a node lies o the shortest path between other nodes
- Closeness Centrality: Is the score assigned to each node relative to its closeness to other nodes
- Eigen Centrality: Is the score assigned to the node’s influence to the number of connection to other nodes. This tells us the influence of a node on a network
- Page Rank: Is the score assigned to the node’s influence to the number of connection to other nodes also taking the weight into account
- Distance / shortest path: How a node travels through a path
Social Network Analysis Example : Call Data Records
Call Data Records are all the details of a particular call. It usually contains the following data:
- Incoming Call Number
- Outgoing Call Number
- Location (Longitude, Latitude)
- Time of Call
- IMSI (SIM Serial Number)
- IMEI (Phone Serial Number)
- Call Details (Voice, Text)
Case Study Details:
If we have extra information for instance IMEI or ISMI number of the incoming or outgoing calls we can find other information within the data like how many times a person changes their SIM card in a phone or the number of phones they change keeping a single SIM. By analyzing the Call Data Records we can uncover patterns of the caller like the peak times they call, the location (latitude, longitude) they call from, who they call and if they are part of a cluster. If we have the actual call details we can use deep learning techniques to assign an ID to the voices on the call. This can help us identify identities of past and future callers. The location of the caller can also help us perform mobility analysis which basically relates to the movements of the caller on a map. Finally when we uncover enough patters we can automatically program the database to uncover these patters by utilizing something called a “database trigger”. Which basically means that once all programmed commands are executed the database automatically produces similar results.