Ai: An Innovative Way to Reshape Quantitative Risk Management
Artificial Intelligence, or Ai, is reshaping industries throughout the world, especially the financial industry.
The capital markets reward new technology generously; Ai and machine learning techniques are now applied to quantitative modeling and investments by some of the most innovative funds to generate significant alpha.
By comparison, Ai and machine learning technology seem to stay in the experimental stage in quantitative risk management, indicating a blue ocean to explore new models and techniques. In what ways can investors make good use of Ai in their risk management, especially when considering the background of the coronavirus crisis and historically high volatility in the US stock market?
It is surprising that Ai has not been significantly applied to risk management. According to Bank of America Research, 80% of publicly available data and information are not structural. Data can take the form of messages and comments people leave in online platforms, including Twitter, with 500 million tweets per day, and Instagram, with 800 million active monthly users.
By applying Natural Language Processing (NLP), a machine learning technique to such non-traditional, unstructured data, risk managers will be able to automatically generate quantitative market sentiment measures that provide innovative insights beyond traditional data.
Furthermore, risk managers can take advantage of the variety of resources online, such as Google’s open-source NLP pre-training system, called Bidirectional Encoder Representations from Transformers (BERT).
Twitter and YouTube also offer their API to developers to mine natural language data in their platform, which makes this technique highly accessible.
Ai and machine learning techniques can also give quantitative risk management an edge by combining new models in Ai with the traditional ones to improve efficiency in cutting costs and better performance.
According to JP Morgan, it is possible to sidestep the Greeks by relying on deep learning and historical data to hedge derivatives and mitigate risk. In this way, many traders believe hedging costs can be slashed by up to 80%. As commented by Hans Buehler, global head of equities analytics, automation, and optimization at JP Morgan, it has been decades since the BS-model governing derivative hedging emerged, and there should be an end to this BS era.
Additionally, machines have the advantages of never tiring and having constant uptime. This significantly promotes risk monitoring coverage while also conserving labor cost.
For instance, in the central limit order book trading, market makers are now able to improve their efficiency to detect fraudulence by Ai and machine learning techniques by over 300%. They are responsible for identifying and preventing fraud, such as traders that set fraudulent orders to manipulate the market’s perception of supply and demand. Because of the complex nature of this problem, it was quite difficult for a person to identify such issues.
Eventus Systems, headquartered in Texas, released their Validus platform to help solve this issue. Validus provides a deep learning model with thousands of parameters that can help identify fraud and other illegal behaviors and determine the priority of the alarms that require the most urgent attention.
Validus uses the findings of its customers' human analysts to continuously train its deep learning models to improve the accuracy of its fraud discovery.
A recent survey conducted by SAS and the Global Association of Risk Professionals found a surging demand for enterprises to increase the adoption of Artificial Intelligence and Machine Learning models to support key risk business use cases.
Enterprises are attracted by the ability to ‘learn’ that enables greater machine learning model accuracy and predictability. This can then be applied to many areas, including credit scoring, risk grading, model validation, and loan provision.
However, there are also some challenges tot Ai and machine learning and their introduction into quantitative risk, the largest being the potential risks associated with it. The first concern is that AI solutions, which learn and evolve over time and contain many hidden decision processing layers, can make auditability and traceability particularly challenging.
Though efficient and effective in practice, machine learning models face such “black box” issues in many scenarios”. Second, the financial industry is highly regulated. It has many business lines and products and always needs to be sufficiently compliant to the regulatory requirements.
The history of regulatory penalties for non-compliance or misconduct in the financial industry introduces an additional level of conservatism in the adoption of relatively unknown technologies in regulated activities.
The overwhelming advantages of Ai and machine learning will undoubtedly reshape the financial industry, including new and innovative applications in the quantitative risk management area.
Written by Nuoyao Yang & Edited by Alexander Fleiss & Gihyen Eom