- Moody’s recent research with Rigetti uses quantum-based signature kernels to predict the likelihood of recessions and to assess how economic and financial data tell us about the risk of future recessions, yielding promising initial results in terms of accuracy and early recession warning capability.
- Ricardo Garcia of Moody’s Analytics said that the rare nature of recessions makes prediction a challenging task because most statistical models rely heavily on examples of past recessions to predict the future.
- Garcia went on to say that they found that the quantum-based signature kernel method yielded promising initial results in terms of accuracy and recession early warning capabilities, demonstrating their potential value for time series modeling.
PRESS RELEASE — May 10, 2023 — With noisy quantum devices becoming more readily available in the near future and the race for fault-tolerant quantum computers in full swing, quantum technology continues to advance rapidly, and it’s becoming increasingly important to understand which applications can benefit from the power of these devices This. Financial institutions have a key role to play in developing cutting-edge techniques that combine quantum computing and machine learning to increase efficiency, reduce risk and deliver better results to customers.
Moody’s recent research with Rigetti using a quantum-based signature kernel to predict the likelihood of a recession and to assess how economic and financial data tell us about the risk of a future recession, yielded promising initial results in terms of accuracy and recession early warning capability.
Quantum computing is an evolving technology, but many quantum hardware architectures are steadily evolving towards higher qubit regimes and lower error rates where QML models can be checked to outpace simulations. Other recent examples of QML applied to finance include leveraging fraud detection algorithms and quantum generative modeling to generate high quality synthetic data for testing asset allocation and risk management strategies.
As the field develops and more tools become available, economists and other practitioners/analysts can begin to leverage this for time series forecasting challenges and other regression-based, clustering and unsupervised machine learning approaches.
Ricardo Garcia of Moody’s Analytics, said: “The rare nature of recessions makes predicting a challenging task because most statistical models rely heavily on examples of past recessions to predict the future. We compare the capabilities of quantum and classical prediction methods in this particular task, using quantum computing to ‘zoom’ the machine learning process. We found it quantum based signature kernel method yielded promising preliminary results in terms of accuracy and recession early warning capability, demonstrating its potential value for time series modeling. Further work will assess its effectiveness in dealing with high-dimensional and irregularly spaced financial/macroeconomic time series and continue to investigate improvements to the quantum model and, more importantly, we will optimize execution time and quantum error mitigation techniques, to outperform classical quantum circuit simulation. on issues with higher feature counts.
“We hope to generate further advances in Quantum Machine Learning (QML) techniques applied to financial and economic time series problems by introducing these algorithmic advances and real-world use cases. Nonetheless, applying QML to classical data sets still remains a major challenge as further empirical evidence and algorithmic designs are needed to demonstrate the superiority of quantum in prediction tasks.”
Complete research can be downloaded at the link https://www.moodys.com/web/en/us/about/what-we-do/quantum-computing/recession-prediction.html