Big Data, Machine Learning, and Artificial Intelligence in Central Banks

With the rapid growth of real-time economic data, the Bank for International Settlements (BIS) reviewed the use of big data and machine learning in central banking in a 2021 paper, which found that:

• Most central banks have programs in place to process applications in:
  • Economic research
  • Monetary policy
  • Financial stability
• Big data is used in supervisory and regulatory processes
• Major challenges include:
  • Data quality, volume, and privacy concerns
  • Ability to provision adequate IT infrastructure
  • Developing necessary human capital

How Central Banks Use MATLAB and Other Products from MathWorks?

MATLAB^® lets you handle a variety of datasets, from running billions of rows on a single machine with tall arrays to parallelizing thousands of cores on Enterprise or cloud data centers with MATLAB Parallel Server™. MATLAB Parallel Server also lets you leverage centralized specialized hardware, like clusters of GPUs, from your desktop without requiring any extra coding.

The ECBs (STAMP€), for example, use the capabilities of Parallel Computing Toolbox™ to run tens of thousands of simulations in a few minutes.

MATLAB Online Server™ enhances scalability and data security by keeping data in a secure cloud environment and bringing compute to the data. Browser-based access to the MATLAB environment maximizes efficiency in the use of compute resources and software administration.

Developing Necessary Human Capital

Using MATLAB and no-code, graphical workflows with apps like the Classification Learner, lets you apply the latest data science techniques in a familiar environment without having to learn new tools. MATLAB, Simulink, and other add-on products provide many domain specific examples that apply machine learning techniques to traditional credit problems such as credit default and stress testing.

Develop relevant skills with resources from MathWorks^® such as comprehensive documentation, video demonstrations of common workflows, and self-paced online training as well as massive open online course (MOOC) selections like Practical Data Science with MATLAB Specialization from Coursera. Additionally, instructors from MathWorks can deliver tailored courses to your economists and data scientists, virtually or onsite, to help them apply these tools to their use cases.

Explaining Machine Learning Models

As the BIS identified in the 2021 working paper, artificial intelligence (AI), machine learning, and data science have become strategic initiatives across Central Banks, creating an environment of rapid change and great potential for innovation. Given the importance of being able to explain complex machine-learning models to senior stakeholders, MATLAB provides several tools for interpreting AI and machine learning models, such as Partial Dependency Plots and Shapley Values, shown in this video, Explainability in the Age of Regulation (56:30).