Machine learning is actively used to solve many problems today. Take for example, the text messaging function on your iPhone. It uses machine learning to try to predict the next word in the sequence. Google Translate applies machine learning to translate between languages. I can cite countless similar examples of application, which we use on everyday basis, which have machine learning at their core.
When it comes to finance, the use of machine learning is much lower. The difficulty of trying to predict financial markets is that they are constantly evolving. The market today is very different from that in 2008. This contrasts to problems which I’ve discussed above. Yes, old words go out of fashion and new words are invented, but the problem is largely unchanged. There is also the importance in finance of being able to interpret why a model has given you a certain answer. However, despite this, there are several areas where we might see machine learning being used more in finance, which we discuss below.
Techniques such as deep learning need a large initial dataset to train the algorithm. In finance, an obvious area where there is a huge amount of data generated is in high frequency trading, where we can collect not only trade data, but also quotes from across the order book. There are notable challenges however, such as the need to execute code quickly, which can be at odds with using computationally expensive models.
Identifying suspicious trading activity for compliance teams can be a challenging problem given the sheer volume of trades executed. Machine learning can be used to identify such activity in an automated manner by going through trading data. It can also be extended to scan traders’ communications, to flag messages for compliance teams to read for any unusual behavior. For consumers, machine learning techniques can be used to help identify potentially fraudulent activity.
Whenever a bank gives loans to consumers, they need to compute their credit scores. Machine learning can be used to scan through their previous spending patterns and general circumstances (such as their job, age and so on) to help establish the likelihood that they’ll be able to pay back the loan.
Big Data could also be helpful for determining sovereign ratings thanks to its ability to process economic data.
The abundance of Big Data about individuals’ behavior can be used to improve economic forecasts. We can for example, use data derived from social media to enhance unemployment forecasts, or
use footfall traffic on high streets to help estimate retail sales. For emerging markets, where economic data might be sparser, we can use machine learning to identify night time lightintensity from satellite photos to estimate economy activity.
It might not be the most glamourous area, but the first step of any sort of data analysis requires an element of data cleaning and finance is no exception. It can be an extremely time-consuming process. We can use machine learning to identify invalid observations in our dataset and free up time for actually analyzing the dataset.
Saeed Amen is a systematic FX trader, running a proprietary trading book trading liquid G10 FX, since 2013. He developed systematic trading strategies at major investment banks including Lehman Brothers and Nomura, and runs Cuemacro, a consulting and research firm focused on systematic trading.
Have a confidential story, tip, or comment you’d like to share? Contact: firstname.lastname@example.org
Bear with us if you leave a comment at the bottom of this article: all our comments are moderated by human beings. Sometimes these humans might be asleep, or away from their desks, so it may take a while for your comment to appear. Eventually it will – unless it’s offensive or libelous (in which case it won’t.)
Image credit: Pogonic, Getty