Understanding the Health of Financial Markets Through Semantic Technology
Understanding the interconnection of data through graph and semantic technology will allow us to detect complex problems and enable new use cases.
- By Carl Reed
- June 27, 2017
During last decade's financial crisis, industry experts perceived various issues emanating from multiple financial institutions, but they were not necessarily able to connect the dots between them until it was too late. Because of the number of interconnections in the world, its markets, and the organizations participating in them, the links leading to financial disaster were too complicated for any but a handful of skeptical investors to notice.
Until recently, there has been no formal solution available to connect these financial data points, which are only increasing in complexity. Today, however, semantic technology can help address these shortcomings, allowing any and all relationships to be represented as distinct entities that can be directly queried and connected using simple graphs.
The objective of semantic technology is to create an enterprise asset of consistent, connected, reusable data that can be created once and leveraged many times to meet regulatory requirements, reduce operational expenditure, and monetize analytics for business value. Implementation is both technically and organizationally complex, however. It must effectively satisfy multiple internal and external pressures (particularly critical in finance) that compete for priority and funding, including:
- Regulatory Compliance: Demands for demonstrable lineage involve source registries, business glossaries, and consistent practices eluding most organizations.
- Operational Efficiency: With expenditures constantly scrutinized alongside projected ROIs, expectations for new operations -- and their efficiency -- become disproportionately high.
- Cybersecurity: Organizations require clean, unambiguous, representative enterprise data signals as a foundation for advanced surveillance-style analytics.
- Business Intelligence: Businesses need uncontaminated data as the foundation for competitive client, market, operational, risk, and reputation advantages.
The traditional response to these pressures is for different business units to create their own locally managed data as byproducts of their respective business processes. Simply replacing the resulting surplus of data marts with cheaper Hadoop implementations does not address the problem or satisfy the emerging regulatory and competitive frameworks for consistent enterprise-quality data.
A better solution is common enterprise semantics supported by a credible data architecture. These semantics should formalize the consistent specification of entities and entity relationships, and they should be able to integrate with any existing and/or new enterprise data ecosystem. Such semantics will drive consistent data as an enterprise asset across the organization.
Different sub-organizations have specific priorities for semantics. Within the financial sector these can be categorized broadly as analytical objectives within client, market, operational, and risk and reputation departments. Traditionally, the separation of these analytical objectives has driven a parochial and disjointed data strategy.
Silos, with their focus on business process, should and will continue to exist. However, with the addition of semantics, silos will be producers and/or consumers of enterprise data that is governed by the enterprise and distributed consistently as an enterprise asset.
Top-down enterprise governance will replace the duplicity and ambiguity of disjointed, narrow data governance. As the consistency of data semantics' process and implementation increases, internal and external trust increases, top-down governance increases, operational efficiency increases, and limitless use cases become applicable.
Empowering Unlimited Use Cases
Although relationship awareness is crucial to exploiting any data set, this utility becomes even more important within financial circles.
That's where semantic graph technology plays a role. The semantic conceptualization and precise definition of the relationships between things enables link-style analytics for insider trading, counter party risk, or supply and demand chains. Graph technology allows these relationships to be formally represented as entities that can be directly queried, traversed, and "reasoned" across.
Any collection of formally defined "things," Internet-connected (IoT) or otherwise, requires a formal understanding of the relationships between things. Semantic graph allows you to implement a representation based on your current interest or understanding and then evolve it over time.
For example, semantic graph technologies enable organizations to model any aspect of insider trading, including electronic communication surveillance, organizational structure, employee-entity relationships, job data, and relevant geospatial and temporal data. The links among all of these data types are necessary to determine the relationship between a particular trader and a specific security, so organizations can effectively monitor the possible movement of material non-public information.
The greater merit of graph-aware analytics in this respect is illustrating how someone could have gleaned information he or she shouldn't have to conduct insider trading, i.e., how an insider might have communicated with a trader. This model provides the basis for an employee-entity graph in which trades, insiders, and traders are all linked, allowing organizations to see the relationships between them and how they may have contributed to insider trading.
The true value in the graph approach for the insider trading use case is found in its recurring relevance for additional use cases. The aforementioned graph is not only critical to regulatory compliance for insider trading but also useful for implementing security. Furthermore, it helps identify viable employees for particular client interactions, cross-reference organizational communication efficiency, and effectively provide a foundation for current or future relationship analytics.
Graph-aware analytics are also ideal for improving efficiency in operations, especially in data centers. The relationships between things and the things of importance are now assets such as switches, routers, subnetworks, networks, machines, virtual machines, applications, business functions, and consuming businesses. The link analysis can proactively define the operational risk that needs to be managed or reactively define what needs to be triaged when something breaks. This same graph can be enriched to represent supply and demand for more effective billing and provisioning.
Another common use case for semantic graph analytics in finance pertains to market intelligence and understanding what forces are impacting a particular market and how. Such graphs are also well-suited to represent supply and demand chains between industry, sector, security, and Global Information Classification Standard (GICS) enhancing market intelligence. This technology is equally capable for cluster analysis, such as collateral concentrations by region, legal entity, etc.
An Enterprise Asset
The application of common semantics to harmonize enterprise-quality data renders data a reusable enterprise asset empowering any assortment of use cases. Semantic graphs and associated graph analytics are emerging as an important information structure.
However, harmonized data is of equal importance across any information structure --from traditional relational data marts to advanced machine learning. It enables organizational reporting, data modeling, and data science to focus on implementing quality solutions to business requirements, instead of spending inordinate amounts of time and money on inefficient and disjointed data engineering.
Carl Reed is formerly a technology managing director of Credit Suisse, serving as the head of data architecture and semantic technology. Carl recently became an industry advisory council member to Cambridge Semantics, a provider of graph-based smart data management and exploratory analytics solutions.