I'm interested in machine learning in relations to the stock market (predicting future values of stocks etc).

What topics would I need to learn - e.g. what branch of AI to look into etc? and what libraries/tools do I need?

  • I think this question could be better answered at the quant site: quant.stackexchange.com – Tangurena Dec 27 '11 at 23:47
  • 2
    @Tangurena: the question is off-topic for Quant SE, as the site is dedicated to professionals only. It would be closed. – SRKX Dec 28 '11 at 0:21
  • 1
    Read "Fooled By Randomness" or "Black Swan" and fughed aboud it! Too bad Mandelbrot died. Predicting the stock market or future in general is not possible. youtube.com/watch?v=QCsLTPGpEfw – Job Dec 28 '11 at 5:24
  • Maybe this question is more suitable for the stats SE. In addition, what kind of master did you do, and how much math did that involve? I did a phd in data assimilation and geostatistics, and I'm daunted by the ML math I see at colleagues. – Paul Hiemstra Dec 28 '11 at 10:16

Machine learning is a whole field in Computer Science, which is quite different from Artificial Intelligence.

Stanford provides a free online class for machine learning which will get you started.

You need to understand that machine learning alone will most probably not be sufficient to create good trading strategies; you need to understand well how the markets work and most systematic strategies require large enough capital to minimize the transaction costs (otherwise the strategy is not profitable, see implementation shortfall).

To be able to understand and apply machine learning, you'll also need a good mathematical background (the better it is, the deeper you can go with the algorithms). I would recommend Information Theory, Inference and Learning Algorithms (MacKay) as a reference book for the field.

| improve this answer | |

I'm working closely with machine learning experts, and I see that it requires a lot of expertise and math knowledge to apply ML correctly, You have not stated what your current background is. If this is mathematics, phyics, or aome other math heavy study, you could probably learn a lot online and from books. If not, to get up to the level to use ML as more than a black box, consider doing a masters degree which incorporates ML. People in finance not knowing statistics/ML, but still applying it because it gets them something that impresses their boss/client, but also could lead them to providing advive which results in the loss of millions of dollars.

I realize that doing a masters degree of two years is quite an investment. However, I think to really use ML sensibly, this could be a requirement.

| improve this answer | |
  • I removed his personal background ("programmer") from the question to avoid it being to localized... Good answer though. – SRKX Dec 28 '11 at 9:34
  • Thanks for the compliment. Programmer is still a bit broad, this can imply everything from a general programmer to a computer scientist. Maybe this question is more appropriate at the stats or math SE. – Paul Hiemstra Dec 28 '11 at 10:14
  • This would require ML and domain knowledge; perhaps an MBA with the appropriate specializations would be a good idea. For all the derision of people who rely on their MBA and don't learn their business, it can teach you an awful lot about the business world. – David Thornley Dec 28 '11 at 17:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.