The Data Exchange Podcast: Murat Özbayoğlu on applications of deep learning to investing and asset management.
In this episode of the Data Exchange I speak with Murat Özbayoğlu, Chair of Artificial Intelligence Engineering at TOBB University of Economics and Technology in Ankara, Turkey. I wanted Murat on to discuss two survey papers he and his colleagues wrote on the use of deep learning in finance:
- Financial Time Series Forecasting with Deep Learning : A Systematic Literature Review: 2005-2019
- Deep Learning for Financial Applications : A Survey
I’ve long been fascinated with finance and trading. My first job after I left academia was as the lead quant in a hedge fund, and ever since, I’ve tried to stay abreast of what tools and techniques quants and data scientists in finance are using. Forecasting in this setting usually means price prediction or price movement (trend) prediction. Output of forecasting models are used to inform investment decisions. What makes finance particularly challenging is that many people are using the same underlying data (time series of prices/values), and thus as Murat notes, many firms use alternative data sources (such as text) as potential sources of additional signal.
Murat and his colleagues systematically go through the literature and they assembled thirteen tables summarizing research into deep learning in key areas of investing and asset management:
Stock Price Forecasting Using Only Raw Time Series Data
Stock Price Forecasting Using Various Data
Stock Price Forecasting Using Text Mining Techniques for Feature Extraction
Index Forecasting Using Only Raw Time Series Data
Index Forecasting Using Various Data
Commodity Price Forecasting
Forex Price Forecasting
Cryptocurrency Price Prediction
Trend Forecasting Using Only Raw Time Series Data
Trend Forecasting Using Technical Indicators & Price Data & Fundamental Data
Trend Forecasting Using Text Mining Techniques
Trend Forecasting Using Various Data
A major caveat is that finance people are secretive and it’s unlikely that they will publish their most effective strategies. Murat reminded me of a joke from finance: “If it’s interesting we’ll keep it to ourselves, if it’s not interesting we’ll publish it!”
Nonetheless, understanding how deep learning can be used for forecasting time series – not just financial time series – is important. We also discuss how organizations should be using machine learning for time series forecasting (as part of ensembles), and the pros and cons of various forecasting techniques.
As Murat notes, currently these deep learning models are best suited for models that are trained offline so realtime applications like high-frequency trading aren’t common:
- “If you want to apply these deep learning techniques in realtime to high-frequency data and you want to train online, that’s a big challenge. … Most of the studies involve offline training … triggering a buy signal and holding for a few days … For such a model you don’t need as much speed, you have all the time in the world.”
Subscribe to our Newsletter:
We also publish a popular newsletter where we share highlights from recent episodes, trends in AI / machine learning / data, and a collection of recommendations.
- A video version of this conversation is available in our YouTube channel.
- Arun Verma: “How machine learning is being used in quantitative finance”
- 2020 NLP Survey Report
- Weifeng Zhong: “Using machine learning to detect shifts in government policy”
- Max Pumperla: “Connecting Reinforcement Learning to Simulation Software”
- One Simple Chart: which sectors are using reinforcement learning
- Alan Nichol: “Best practices for building conversational AI applications”
- Denise Gosnell: “How graph technologies are being used to solve complex business problems”
- Edmon Begoli: “Hyperscaling natural language processing”