The challenge:
Build a system with the Swiss National Bank (SNB) to analyse the interbank transactions and detect substantial payment difficulties of participants that are crucial to the Swiss interbank Clearing (SIC) payment system.
The result:
Development and delivery of a customisable AI-powered monitoring tool with data visualisation and highly accurate nowcasting capabilities. The tool leverages a novel five-step supervised Machine Learning methodology to analyse big data sets, identify outliers, and diagnose network participant stability.
swissQuant’s Machine Learning models transform Big Data sets into actionable insights
Key solution elements:
- Robust data processing: co-development of ‘stress event’ labels as well as a trend and seasonally adjusted data set based on aggregated raw data (CDS spreads and SIC payment transactions).
- Outlier identification: individual outlier scores calculated via application of our heterogeneous outlier ensemble Machine Learning model.
- Layered Machine Learning model: the likelihood of stress events at banks derived by employing a supervised Machine Learning approach to outlier scores.
- Predictive performance: evaluated via application of ROC curve metrics, qualitative analysis, and visualisation techniques.
Strengthening data-driven decision-making for Central Banks and beyond:
Application of swissQuant’s methodology and modelling led to very accurate stress event nowcasting for banks within the SIC payment system. This novel methodology and its specific application were captured in an academic paper jointly published by swissQuant and the SNB, with the findings presented at scientific conferences in Switzerland and abroad.
Both the private and public sectors are engaged in high-level conversations regarding the feasibility and necessity for tools that are grounded in timely data and that can be used in policy-making processes. swissQuant is proud to offer best-in-class products and services that empower leading institutions to optimise their processes by leveraging emerging technologies and sophisticated quantitative analysis.
Download the PDF:
Click here to read the original article.
