Adopting Artificial Intelligence in Finance

CFA Institute’s Larry Cao discusses the opportunities and challenges of adopting AI tech in the investment industry.

If there is anything that will surely stay on the radar of the finance industry for some time, it ought to be the development of advanced technologies. Big data and artificial intelligence emerged long ago to analyse financial data and create financial models, yet it is only recently that their application and impact has become more widely understood.

Artificial intelligence (AI), the study of emulating human understanding and reactions to different circumstances, can range from the virtual assistant installed on your phone all the way to deep learning neural networks that mimic the human brain. To equip AI with the ability to self-learn instead of humans constantly adapting codes and inputting data to assign actions, “machine learning” (ML) is an area of AI that we turn to.

Those who tried using the popular chatbot ChatGPT should not be a stranger to machine learning, as one of the backbone technologies behind it is “natural language processing” (NLP), which falls under the umbrella of machine learning. NLP enables computers to study the interpretation of textual and spoken data, and with cross-disciplinary input, generate models that could enable machines to understand words like we do.

Machine learning in finance

Machine learning is certainly useful for data scientists and investment professionals in many aspects. Some may use ML to predict future stock prices, in order to make more informed investment decisions and increase potential returns. Others may use ML to give a forecast on the securities that will likely hit rock bottom in the near future, which could be exceptionally useful during a turbulent market.

However, anyone who has tried to research the stock market would understand the drastic difficulties of making accurate predictions as there are simply too many variables at play, some of which (such as human emotion) are not able to be processed by machines. An easier task would therefore be the prediction of fundamental company variables like corporate earnings, which can help analysts to build a more holistic view of the prospects for a particular company.

Apart from the predictive function, ML could also be used to solve language problems with the assistance of NLP. A fairly common issue for investors around the world is language barrier. Despite the many talented professionals in the industry who speak more than one language, financial markets tend to generate colloquial terms that are exclusive to each language. In this case, we could leverage the capabilities of ML and NLP to search for explanations of specific jargons.

As powerful as they seem, ML algorithms face certain difficulties in financial markets and major limitations exist when we apply this technology. Financial data is notorious for its low signal-to-noise ratio, meaning there is generally no dominant variable for the performance of any given security. Algorithms might not be able to correctly distinguish the underlying relationship between the variables and outcomes of the market beyond the sample period on which the algorithm was trained.

On top of this is the amount of available financial data, which far exceeds other areas where ML is being utilized such as the consumer internet. The result is that ML algorithms may have perfect hindsight but poor foresight. Last on the list is the adaptive and irrational nature of the market. Unlike other static systems like those found in the sciences, behind all the numbers and dollar signs in the financial markets lie the human mind, which is often irrational and highly unpredictable. These behavioural factors can confound algorithms.

What about Natural Language Processing

Natural language processing has a wide range of applications in the financial industry. Investment researchers and analysts can add NLP to their suite of media tracking tools, which extract and analyse keywords from text and audio data. These tools are used to track public opinions and sentiments on social media, which can be a key influence on the stock market, as well as parse financial information in company earnings calls.

Investment professionals can also benefit from the “topic modelling” function. For instance, financial industry practitioners can discern key themes within a collection of statements or press releases from a central bank to better predict policy changes. On the other hand, NLP is also useful when it comes to discovering risks in corporate filings. By parsing the filings over a long time frame, text mining can dig out major differences or inconsistencies among the documents, rendering the technology useful for risk management and compliance purposes.

In order to best utilize NLP, there are a few key challenges that we should address. First, there exists some technical hurdles for the technology, most commonly in dividing up sentences and tagging the parts of speech. These hurdles also vary among different languages which have different syntax and semantics. To tackle this issue, dictionaries are particularly helpful and can improve language learning in specialized domains. Moreover, accessibility of data is a crucial part (again) and it is encouraging to see more open-source tools and databases being made available to researchers.

Finally, at the level of the firm, adopting advanced technology could mean a shakeup for the whole organization. It will require a strategy and culture conducive to experimentation and cross-functional collaboration in what we call “T-shaped teams” of data scientists and investment professionals, each with opposite and complementary skills.

Where to start

For those who are taking their first step on the journey of AI adoption, it is normal to feel a bit lost. Appointing a key person who is experienced and knowledgeable in the relevant area is the best place to start. Some investment analysts who have a passion for advanced technologies would be a great addition to the team’s capabilities. In the upper hierarchy, support from the senior leadership is just as important. They do not have to be experts in ML or NLP, but they should understand that these new technologies rely heavily on trial-and-error, so realistic expectations can be set.

Similar to any other new projects and endeavours, companies are often concerned about the performance of these technology adoption projects. Return on investment is likely to be a prioritised consideration, but we do recommend taking a step back and looking at the bigger picture. For instance, does the new approach generate unique insights in a more efficient way? And more importantly, can you see scalability for the endeavour to be applied across more datasets or applications with similar traits?

The realm of AI is still rapidly developing and has great potential to be more widely used across many industries. Investment professionals ought to keep a close eye on the latest developments and remain open to them. Stay informed to stay ahead.

Larry Cao is Senior Director of Industry Research at CFA Institute, and author of the Handbook of Artificial Intelligence and Big Data Applications in Investments.

To Top
Share via
Copy link
Powered by Social Snap