As firms strive for competitive prowess, artificial intelligence (AI) and machine learning are increasingly scaled across multiple business areas in financial services, a recent Refinitiv report reveals. With financial data scientists adopting a more influential role driving machine learning strategies, firms are only limited by challenges around data quality and availability, writes Amanda West, global head of Refinitiv Labs
- In its new report, Refinitiv captured insights from over 400 data scientists, quants and data leaders to reveal the latest AI and machine learning trends in finance.
- Machine learning is clearly maturing in financial services. Eighty percent of firms are making significant investments in related technologies.
- Data scientists play an increasingly strategic role, and the number of data science teams in financial services firms have grown by more than 260 percent since 2018.
These are some of the highlights from the second Refinitiv global Artificial Intelligence and Machine Learning Report, entitled ‘The rise of the data scientist’.
The report is based on one of the largest global surveys of data practitioners and decision makers on the use of machine learning across finance, and is a ‘must-read’ for data practitioners and innovation leaders.
Watch: New AI and Machine Learning Research—The Rise of the Data Scientist
Machine learning becomes a horizontal capability
Machine learning is maturing in financial services, as companies deploy ever more sophisticated techniques, such as deep learning, and begin to execute rapid innovation cycles.
Over 72% of this year’s survey participants say it is a core component of their business strategy, with 80% making significant investments in associated technologies.
Challenges relating to investment levels, technology choices and access to talent, identified as key barriers to adoption in 2018, have diminished, providing a robust foundation to implement machine learning models at scale.
Data scientists have a more strategic role
Data scientists have shifted from the technology department to working and driving machine learning directly in the business. They have transitioned from developing models at the request of the business, to influencing the technology and data strategies required to achieve business objectives.
The importance of data scientists to the success of machine learning strategies is clear. The average number of data science roles in each firm has increased significantly since our first machine learning report in 2018, and the number of teams has risen by over 260 percent.
Adoption of deep learning
This year’s report shows that 75% of firms are employing deep learning. An unexpected advance, but backed up by a surge in the use of leading deep learning technology frameworks.
Deep learning can help firms gain valuable insights from large and diverse unstructured data sets, but the cost of hardware and time taken to train models can be prohibitive.
Natural language processing (NLP) is another growing area of focus, and is increasingly applied to unlock value in unstructured data. It’s no surprise that the use of unstructured data is on the rise. In 2018, only 2% of firms worked exclusively with unstructured data, and this figure has shot up to 17% in 2020. Today, 64% of firms work with some form of unstructured data.
The importance of data strategy
Data scientists are using diverse data sources, including unstructured data and alternative data, to develop models.
Only 3% of participants in the 2020 survey say they do not use alternative data in their models, a decrease from 30% in 2018, which reflects the need for new data sets to better understand the impact of Covid-19.
In our 2018 survey, poor data quality was considered one of the key challenges to the effective deployment of models. This year, poor data quality and availability continue to be cited as barriers to adoption.
If firms are going to genuinely benefit from the speed, agility and value of an ‘AI-first’ vision, they need high-quality, trustworthy data that can be easily accessed, ingested, and manipulated.
Covid-19 is the arch accelerator
Our 2020 report shows that 72 percent of firms’ models were negatively impacted by Covid-19. Some firms declared their models obsolete, while others are building new ones.
The crux of the problem was a lack of agility to quickly adapt and incorporate new data sets as circumstances changed.
What you need to know about machine learning in finance
This year’s report is packed with everything you need to know about the machine learning landscape. It provides insights into use cases in financial services, the changing role of the data scientist, the use of emerging technologies, the latest investment trends, and more.
It also includes practical advice based on Refinitiv Labs’ experience in developing and deploying machine learning models to solve some of the most challenging problems in finance, and shares some of the prototypes built in collaboration with our customers.