Finance, like many other industries, is exposed to social interactions. Such a nature of the domain requires processing an enormous amount of data and building a system resistant to such requirements. Therefore, the foundations of technology used in the beginnings of the financial industry are still used today. For the same reason, this industry is not subject to agile and large-scale changes.
However, some of these changes are becoming ubiquitous nowadays. Thus, for example, by modeling the previously mentioned interactions, graphs enter the world of the financial industry.
By analyzing the connections between users and resources, we can get an insight that we would not have the opportunity to see using standard technologies such as relational data containers. In addition, the advantage of graph data is the ability to faster analyze and use the tools available particularly for graphs.
Within the financial industry, special emphasis is placed on data security. Thus, it is necessary to monitor when and under what conditions the data were moved from one system to another. Data lineage is taken quite seriously.
Therefore, implementation via graphs makes it easier to monitor data, not only within the storage structure but also visually. Using the shortest path, it is possible to track how a resource has behaved over time and see which channels it passed through before reaching the current one.
Fraud is an unwanted side effect in financial systems and needs to be prevented. Due to the high number of processed data per day, it is difficult to catch which of them are suspicious and which are not. Fraudsters use methods that are based on certain patterns of behavior. Such patterns can be identified using graphs.
Transaction data is easy to model as a directional graph that knows the sender and receiver. Such a system is simple in structure and can be used to detect potential fraud. Users otherwise behave according to similar principles and patterns of behavior. However, when any deviation from this behavior is noticed, for example, a closed circle of recently created accounts that transfer money, it needs to be detected and reported.
Modeled structures in the insurance industry are much more complex than in common finance. These include damages, insured persons, accomplices in the accident, contacts, telephone numbers, medical institutions, their employees, etc.
What can help us greatly are investigation units that have previously encountered scams and have marked previously caught scammers. The magnitude of such scams can be in the millions.
By using features such as closeness to previous scams, centrality, community detection, and many others, it is possible to make inferences over newly arrived data and thus discover whether a new data point is a fraud. Such models are sensitive to change and often miss in their judgment. That is why it is important to have a team of trained people who can make an educational assessment on such cases.
Money launderers are becoming more sophisticated every year, making them difficult to track. Manual investigation on such cases can take up years. Money laundering is detected via suspicious payment chains. These structures are easy to investigate and operate with graphs, and therefore making them the most efficient technology to tackle this kind of fraud.
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.