Algorithms based on artificial intelligence shows incredible promise in making profitable investments.
A recent study by researchers at Germany’s Friedrich-Alexander-Universität (FAU) School of Business and Economics explored the potential of Deep Learning in finance.
They applied the AI-based algorithms to the S&P 500 constituents from 1992 to 2015. The investments generated annual returns in the double digits, with the highest profits made at times of severe market volatility and turmoil.
These algorithms were inspired by the structure and function of the brain called artificial neural networks, capable of learning from examples and independently extracting relationships from millions of data points.
‘Artificial neural networks are primarily applied to problems, where solutions cannot be formulated with explicit rules,’ explains Dr. Christopher Krauss (of the Chair of Statistics and Econometrics, FAU).
‘Image and speech recognition are typical fields of application, such as Apple’s Siri. But the relevance of deep learning is also increasing in other domains, such as weather forecasting or the prediction of economic developments.’
Krauss and his team were ‘the first academics to apply a selection of state-of-the-art techniques of artificial intelligence research to a large-scale set of capital market data.’
‘Equity markets exhibit complex, often non-linear dependencies,’ says Dr. Krauss.
‘However, when it comes to selecting stocks, established methods are mainly modelling simple relationships.
‘For example, the momentum effect only focuses on a stock’s return over the past months and assumes a continuation of that performance in the months to come. We saw potential for improvements.’
One of the team’s models returned 73 percent annually from 1992 to 2015 after accounting for transaction costs, easily beating a real market return of 9 percent annually.
The AI stock pickers performed considerably well during times of market turmoil.
When the tech bubble burst in 2000 the AI-based model returned 545 percent, and in 2008 when the global financial crisis was nearing its peak, the model returned 681 percent.
‘Our quantitative algorithms turned out to be particularly effective at such times of high volatility, when emotions dominate the markets,’ said Krauss.
Follow-up projects with larger data sets
Krauss noted that during the last years of the team’s sample period, profitability decreased and even became negative at times.
‘We assume that this decline was driven by the rising influence of artificial intelligence in modern trading – enabled by increasing computing power as well as by the popularisation of machine learning,’ he said.
However, the team believes that deep learning has a lot of potential.
‘Currently, we are working on promising follow-up projects with far larger data sets and very deep network architectures that have been specifically designed for identifying temporal dependencies,’ says Dr. Krauss.
‘First results already show significant improvements of predictional accuracy – also in recent years.’
Christopher Krauss, Xuan Anh Do, Nicolas Huck, (2017) “Deep neural networks, gradient-boosted trees, random forests: Statistical arbitrage on the S&P 500”, European Journal of Operational Research