Over a million developers have joined DZone.
{{announcement.body}}
{{announcement.title}}

The Bigger Picture of Credit Scoring and Analytics: Enterprise Decision Management Systems

DZone's Guide to

The Bigger Picture of Credit Scoring and Analytics: Enterprise Decision Management Systems

Learn about the decisions and analytical processes that go into the bigger picture of determining a credit score through enterprise decision management systems.

Free Resource

Access NoSQL and Big Data through SQL using standard drivers (ODBC, JDBC, ADO.NET). Free Download 

The previous articles in this series described the key elements of credit-scoring toolkit including scorecard model, scoring strategy, implementation, and monitoring. By putting these single pieces together, we start building a bigger picture of the enterprise decision management (EDM) system. However, this is still insufficient for executing the complete credit-risk decision process. Making the full EDM picture requires bringing the additional jigsaw pieces together, including customer application processing, internal and external data gathering, policy rules, additional analytical models for fraud detection and risk management, optimisation, overrides, and so on.

With its three fundamental components — data, logic, and interface — an EDM system provides the framework for translating data into actionable decisions using data-, model-, knowledge-, communication-, and document-driven decision-making processes. A decision management system is only valuable if it can fulfill the following:

  • Automation
  • Data and systems security
  • Concurrency of processes
  • Scalability that is capable of handling a growing amount of processes that are both easy to change and extend
  • Transparency, so technical and non-technical professionals can understand, share, assess and audit business processes
  • Heterogeneity with diversity of data sources, synchronous, and asynchronous processes, both local and remote

Business decisions are the key outputs of an EDM system. Decisions are consumed within a business process flow and can be reused in other processes. A mix of business rules and advanced analytics is typically considered when creating a decision requirements diagram for credit risk (Figure 1).

Figure 1: Example of a simplified decision requirements for loan application process using BPMN

Business rules can be extensive encompassing internal and external policies, regulations and best practices; typical examples of business rules include age requirements, employment status, credit history, bankruptcy, and write-off history, different fraud rules, in-house records on existing products, and similar.

Overrides are a form of business decision that can overrule decisions based on a credit score cutoff value. Overrides can either approve applications that would have been rejected by the score cutoff, or reject applicants who would have been scored above the score cutoff. Reasons for overriding the model scores are based on specific company rules and exclusion criteria.

A number of predictive models can be utilized within a business process flow assessing various risk elements, such as fraud, delinquency, default, and churn, or calculating applicants' affordability and lifetime value. In addition, a number of optimisation levels can be added in the process flow considering different objective functions such as minimising operational cost or maximising the margin. Return on investment (ROI) analysis — measuring the impact of business decisions — can also be the part of the business process, informing the optimal decision strategy.

Model management and monitoring are additional important components of an EDM system. For improved decision making, advanced business processes may incorporate adaptive analytic techniques through utilization of model monitoring capability and machine learning algorithms. Any performance degradation of existing predictive models is captured using the model monitoring capability that, in turn, provides automated feedback into machine learning models enabling self-correction in real-time. This can significantly shorten a change management cycle and improve the effectiveness of the decision-making process.

The complete credit risk decision process can be fully customised and proprietary to a financial institution. This, however, carries many risks, including complex system maintenance, human resources, and high cost. An alternative solution would be to opt-in using a commercial EDM system that would be faster to implement, require fewer resources, and drive cost efficiencies through investment in technology.

Commercial EDM systems are typically equipped with visual programming features and a point-and-click user interface enabling data scientist or business analysts with no programming skills to create decision requirements diagrams, specify input parameters, control model outputs, implement business rules, run processes in parallel or direct outputs to other internal or external processes.

Ease of implementation, speed of change, compliance with regulations and components modularity are some of the great benefits of EDM systems. They are the brain of a complex body that orchestrates its single components playing a harmonic business symphony.

With this, we end the credit scoring series that demonstrated the development journey from end to end. We hope you have found it informative and interesting to read. For credit-risk professionals, this might open up some alternative approaches for further exploration. For novices in the credit-risk arena, this might be a good starting point for a long and often challenging credit-risk management journey but certainly enjoyable and exciting.

If you need more information on this topic or would like to discuss it further, please email us and we will be happy to discuss and get your ideas.

The fastest databases need the fastest drivers - learn how you can leverage CData Drivers for high performance NoSQL & Big Data Access.

Topics:
big data ,data analytics ,credit scoring ,enterprise ,decision management ,tutorial

Published at DZone with permission of Natasha Mashanovich, DZone MVB. See the original article here.

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}