Risk Analysis Based on Community Detection

Artificial Intelligence solutions

analisi del rischio

Risk analysis is now a fundamental step for banks, insurance companies, and financial brokers due to the increasing complexity and rapid evolution of regulations in the financial context. It also represents a significant competitive advantage for these institutions, made possible by Artificial Intelligence (AI) techniques and Big Data Analytics.

Compliance and risk management have become absolute priorities for financial institutions, and regulatory technologies, known as RegTech, are emerging as essential allies to tackle these challenges. In particular, Anti-Money Laundering (AML) analysis stands out, playing a crucial role in preventing illicit activities and protecting the integrity of the global financial system.

This article will explore how risk analysis can be enhanced through the use of community detection, an advanced data science technique that identifies and monitors networks of relationships within financial data.

We will discover how these innovative technologies not only improve the efficiency and effectiveness of compliance operations but also contribute to a safer and more transparent financial landscape. AI techniques can indeed be leveraged to improve the efficiency and accuracy of transaction monitoring, enabling the early identification of anomalous scenarios.

RegTech and AML in Risk Analysis

In the context of risk analysis in banking, the term Regulatory Technology (RegTech) refers to the management of regulatory processes within the financial sector through the use of advanced technologies. Essentially, RegTech supports financial institutions and other regulated companies in addressing and managing the growing regulatory requirements. Here are some of its main functions and utilities:

  • Automation of Compliance Processes: RegTech automates many of the manual activities related to regulatory compliance, reducing human errors and increasing operational efficiency.
  • Real-Time Data Monitoring and Analysis: It allows continuous monitoring of financial transactions and real-time analysis, identifying potential risks and suspicious behaviors that might indicate illicit activities.
  • Cost Reduction: By automating and optimizing compliance processes, RegTech helps reduce the operational costs associated with regulatory management and compliance checks.
  • Risk Management: It enhances companies’ ability to manage regulatory and operational risks using advanced data analytics, machine learning, and artificial intelligence techniques.
  • Adapting to Evolving Regulations: It facilitates rapid and efficient adaptation to constantly changing regulations, ensuring that companies remain compliant with current laws.
  • Reporting and Transparency: It automates report generation for supervisory authorities and improves the transparency of internal processes, easing communication with regulatory bodies.
  • Fraud and Money Laundering Prevention: Specific RegTech tools are designed to prevent fraud and money laundering activities by analyzing large volumes of data to identify suspicious patterns and anomalous behaviors.

RegTech can be divided into three macro-areas:

  1. MonitorTech: Focuses on continuously monitoring financial activities to ensure they comply with current regulations.
  2. LegalTech: Ensures that financial institutions comply with all applicable legal regulations.
  3. ReportTech: Involves the collection, analysis, and reporting of relevant information to regulatory authorities.

A key function of ReportTech is Anti-Money Laundering (AML), which involves identifying anomalous events such as money laundering or fraud that must be reported to competent authorities like FATF-GAFI.

AML through Community Detection in Risk Analysis

Anti-Money Laundering (AML) analysis using community detection for risk analysis leverages advanced data science techniques to identify and monitor suspicious relationship networks within financial data. In a network composed of users, community detection algorithms identify groups of users particularly connected to each other. Let G = (V, E) be a directed graph, where V is the set of users and E is the set of edges between pairs of nodes (i, j) with i, j ∈ V. Within G, strongly connected components can be identified, i.e., groups of nodes where there exists a path for each pair of nodes within the same community.

In the AML context, community detection-based approaches focus on identifying groups of users showing unusual or suspicious behaviors. These behaviors can be highlighted by analyzing their social and financial interactions.

More formally, the network of subjects can be modeled through a directed graph, composed of nodes and edges. The nodes represent the subjects, while the edges represent semantic links between subjects. Specifically, in the presented approach, the edges can be of two types:

  • Social: Represent subjects co-owning the same bank account, account holder-delegate, account holder-client, and so on.
  • Financial: Represent financial transactions between two subjects.

By combining community detection techniques with temporal motifs, it is possible to identify groups of subjects acting in unusual or non-compliant ways. Temporal motifs are patterns of interactions between users that repeat over time. For example, a group of users might exchange money regularly every month.

External data sources, besides banking data, can provide additional clues about potentially suspicious subjects. Integrating lists of politically exposed persons or high-risk countries, for example, can be crucial in characterizing groups of subjects identified as potentially suspicious.


The aforementioned approach to risk analysis offers a notable advantage over traditional methodologies. Specifically, the main points of the strategy are two:

  1. Subjects are connected through heterogeneous edges: Social and financial edges.
  2. Subjects are grouped using community detection techniques: Communities are analyzed using temporal motifs.

In the AML context, traditional approaches typically use predefined thresholds to identify anomalous flows. However, these approaches focus on single or pairs of subjects, and in the example mentioned above, the money cycle would not be captured. Conversely, using an algorithm that considers relationships (both social and financial) between subjects, through the aforementioned components, it is possible to have an overall view instead of focusing on individual subjects.

NDG/Rapporto ComuneIndirizzoSAEAtecoSSE
00000000SalernoVia XXX 11430421412
00000004SalernoVia XXX 11600600
00000005SalernoVia XXX 11600600
Table 1: Exemplary Table SAE = “Subgroup of Economic Activity”, SSE = “Synthetic Economic Sectorization”.
Soggetto 1Soggetto 2Tipo Relazione
000000004000000000titolare effettivo
000000000000000004titolare effettivo
Table 2: Exemplary Table of Social Edges Between Subjects. The columns “Subject 1” and “Subject 2” contain the NDGs corresponding to the subjects involved in that social relationship.

Application Case

Community detection algorithms represent a crucial innovation in AML, allowing for the identification of anomalous behaviors by analyzing user interactions. In Revelis, this approach has been implemented within the Moneying platform, offering, along with other AML solutions already present on the platform, a broader perspective on potentially dangerous transactions. These technologies can make the financial sector safer and more resilient, improving the ability to prevent illegal activities and ensure regulatory compliance.

Thanks to the previously described procedure, it is possible to generate reports for each identified suspicious community. An example of a community is presented in Tables 1, 2, and 3. Specifically, the algorithm identifies the community composed of the subjects described in Table 1. Table 2 lists the social edges between the community subjects, and finally, Table 3 describes the financial edges, characterized by the execution date, analytical cause, and amount of money.

The prepared reports provide the domain expert, who will then make the final decision, with an analytical view of the community. Notably, money cycles between community subjects can be observed, such as between subjects “000000000” and “000000006”. In the presented example, some data have been masked to preserve the privacy of the subjects in question.

Soggetto 1Soggetto 2DataCausale AnaliticaAmmontare
Table 3: Exemplary table of financial arcs between subjects. The columns ”Subject 1” and ”Subject 2” contain the NDGs corresponding to the subjects involved in that social relationship.

Author: Simone Mungari