The amount of data collected each day is rising at an alarming pace. Social media platforms such as Facebook, LinkedIn, Twitter, and many others save interactions between users while also tracking their individual behavior on a daily basis. While this data is essential in making intelligent and business-critical decisions, it is becoming increasingly difficult to analyze such data structures in the search for meaningful insights. The field of social network analysis (SNA), or more broadly known as network science, utilizes graph theory to better understand such structures. Even though the name might suggest otherwise, techniques that are used for social network analysis can be applied to various other network structures such as transportation networks, shipping networks, financial networks, etc.
Advertising in social networks
Social networks are essential to both social scientists interested in how individuals interact and businesses attempting to target customers for advertising. For example, if advertisers connect three people as friends, coworkers, or family members, and two of them buy the advertiser's product, the advertiser may choose to spend more money on advertising to the third hold-out, assuming that this target has a high propensity to buy the advertiser's product.
While humans are very good at detecting distinct or repetitive patterns among a few components, the nature of large, interconnected networks makes it practically impossible to perform such basic tasks manually. Groups of densely connected nodes are easy to spot visually, but more sophisticated methods are needed to perform these tasks programmatically. Community detection algorithms are used to find such groups of densely connected components in various networks.
Get custom recommendation
A very common problem in network analysis is link prediction. The algorithm predicts which new interactions among the network members are likely to occur in the near future. One way of predicting these links is by measuring the “proximity” of nodes in a network. This can be done by using similarity algorithms such as cosine similarity, Jaccard similarity and overlap similarity.
Where to next?
This text is a summary of one area that fits perfectly with the application of graphs. Therefore, we would like to have you with us when implementing some of these solutions. Share opinions, experiences and problems you encounter when working with Memgraph on our Discord server. We are here for you and we will help you along the way.