How to use "Machine Learning" to build trading system?

C

cf0532

nowadays I am studying 'Machine Learning' after got some tutorial course about it; but until now I don't know exactly how to use it for trading stock or future.
Now I want to learn how to use "machine learning" on trading, so shall I get some tutorial courses about it, especially by Matlab and Python, such as some books?
Thank you!
 
Have you already developed and tested a trading system the classical way?
It's always about generating signals for making trading decisions.
Feed the algo with market data and analyse its answer, ie. its signal.
 
cf0532 said:
nowadays I am studying 'Machine Learning' after got some tutorial course about it; but until now I don't know exactly how to use it for trading stock or future.
Now I want to learn how to use "machine learning" on trading, so shall I get some tutorial courses about it, especially by Matlab and Python, such as some books?
Thank you!
More...

Answers on quant-stackexchange: http://quant.stackexchange.com/ques...-machine-learning-algorithms-to-stock-markets

I have to admit, using machine learning (ML) to build trading systems is not my first choice, as I find it hard enough to build a profitable trading system without it :rolleyes: I've enough experience to know there's no advanced/complex magic formula that will always work.

However, I did some searching and this looked promising, if nothing else, should provide a good learning platform as it's used for education already:

http://www.cs.waikato.ac.nz/ml/weka/documentation.html

However, if you have no idea on how to apply this to trading, then what you really need is to investigate trading with ML fresh in mind, and see if a good idea pops up in a year or 10 :cool:

People seem to use ML for many different things and in many different ways, so it's imperative to know the domain, have some concrete goals to reach according to viable ideas and then ML may become a support-tool in any part of the typical trading processes: analysis, decision, execution.

The right way to go about tools is only using them when having an insatiable need for it to accomplish some tangible goal. Ie. for trading this could be finding some sort of edge. If you have no requirements, you have no use for it. Also, beware studying just for studying's sake. When being taught, often you are served made-up problems together with perfect text-books examples. Such activity may passify the curios and inquisitive mind. You can in many ways learn much more from applying, ie. actually getting your feet wet so to speak. Yes, it's more dirty, it's imperfect, but can be made effective and practical. In reality, no course or book on trading, can give you a profitable strategy, only you can create and maintain that for yourself.

First step could be to go more in-depth in both ML and trading and play with it from there. If you stop making progress in one area, you could focus in the other in the meanwhile.

It may also help to partner up with someone or help someone else having :banghead:
 
Last edited:
It's interesting that a whole cottage industry has spawned quant outfits * and products that are generating programs to disseminate an edge from "randomness" ( of short term price movements in the "markets" ). It's a catch 22. If the signals are based on random information and are keying off of other input by other machines, how can a machine ( or human ) learn from it or trade off of it ?

* One of the original code driven, robotic trading outfits was called " The Prediction Company". https://en.wikipedia.org/wiki/Prediction_Company
Are we all too smart for our own good ? http://www.bloomberg.com/news/artic...at-the-gate-the-inexorable-rise-of-diy-quants
 
Stockmarketmap said:
It's interesting that a whole cottage industry has spawned quant outfits * and products that are generating programs to disseminate an edge from "randomness" ( of short term price movements in the "markets" ). It's a catch 22. If the signals are based on random information and are keying off of other input by other machines, how can a machine ( or human ) learn from it or trade off of it ?

* One of the original code driven, robotic trading outfits was called " The Prediction Company". https://en.wikipedia.org/wiki/Prediction_Company
Are we all too smart for our own good ? http://www.bloomberg.com/news/artic...at-the-gate-the-inexorable-rise-of-diy-quants
More...

First question is if the markets are random or chaotic. It is seemingly small but important difference. Even if we would assume it is random this randomness does not always follow pure symmetric distributions. It is imperfect and hopefully non-random. Detecting, learning and applying knowledge using those imperfection is what ML would attempt to find in the sea of chaos.
 
You must log in or register to reply here.
Back
Top