Bank of England Reports that Machine Learning Methods in Financial Services Could Enhance Business Processes

A report from the Bank of England (BoE) notes that the application of machine learning (ML) methods in the financial services sector could enhance routine business processes. 

Improved software and hardware and increasing volumes of data have enhanced the pace of ongoing ML development, BoE’s report mentioned. The UK’s financial industry is taking advantage of ML technology, which improves the overall efficiency of financial services and markets. 

ML-based systems also make financial systems more accessible and helps to tailor them to specific consumer requirements. Existing risks may increase if governance and controls do not use the latest technology, the report noted. It added that ML also raises questions regarding the use of data, automation of business processes, decision-making, and the complexity of techniques.

The BoE and UK’s financial regulator, the Financial Conduct Authority (FCA), have expressed an interest in the way ML is being used by traditional financial institutions. The BoE and FCA performed a survey this year to gain a better understanding of the use of ML in UK’s financial industry. 

Approximately 300 companies participated in the survey, including major banks, credit brokers, e-money service providers, financial infrastructure firms, fund managers, insurance providers, non-bank lending firms and principal trading platforms. A total of 106 responses were obtained.

The survey asked questions about which type of ML had been deployed, the business areas where the technology would be used and the maturity of applications. The survey also obtained information regarding the technical requirements of different ML use cases. These included how testing and validation was performed on various models, the protection built into the software, the types of data analysis and methods used, and assessments made regarding benefits, risks, complexity and governance.

Machine Learning based systems are increasingly being used in UK’s financial industry. Two thirds, or 66%, of those responding to the survey said they currently use some form of ML Click to Tweet

While the survey’s results may not be statistically representative of UK’s financial system, they do offer valuable insights.

As noted by the BoE, the main findings from the survey are as follows: 

  • ML-based systems are increasingly being used in UK’s financial industry. Two thirds, or 66%, of those responding to the survey said they currently use some form of  ML.
  • The median company uses live ML-enabled applications in several business areas and this trend is projected to double in the coming years. 
  • ML development has entered the advanced stages of deployment in certain cases. 
  • One third, or 33%, of ML-based programs are used for many different activities in specific business areas. 
  • ML deployment is most advanced in the banking and insurance industry. 
  • ML is used in front and back-offices across a wide range of business areas. 
  • The technology is also used in anti-money laundering (AML), fraud prevention, and customer-facing applications (customer services and marketing). 
  • Some companies use ML for improving credit risk management, trade pricing and execution, and general insurance pricing and underwriting.
  • Regulation is not considered a barrier, however, some companies emphasize the need for clearer guidance on how to apply current regulatory guidelines. 
  • Companies do not believe regulation will prevent or adversely affect ongoing ML deployment. Legacy IT platforms and data limitations could slow down the adoption of ML-based systems. Companies said that additional guidance on how to apply current regulation could help ML deployment.
  • Firms believe ML does not always create new risks, however, it could amplify existing risks. For example, ML applications might not work properly, which may occur if model validation and governance frameworks do not adopt the latest technology.
  • Companies use safeguards to manage risks associated with ML, including alert systems and “human-in-the-loop” mechanisms. These are useful for flagging if the model does not work properly.
  • Firms validate ML applications before and after the are deployed. Validation methods include outcome-focused monitoring and testing against benchmarks.
  • Many companies say that ML validation frameworks must evolve as ML applications begin to scale and become increasingly complex.
  • Companies usually develop ML applications in-house. They may also rely on third-parties for the underlying platforms and infrastructure (e.g. cloud computing).
  • Most of the users apply existing model risk management frameworks to ML applications. However, many note that these frameworks must evolve as ML techniques have become more advanced. 

The BoE and the FCA are planning to create a public-private group to address some of the questions and technical areas discussed in this report.





Sponsored Links by DQ Promote

 

 

Send this to a friend