Win a copy of Transfer Learning for Natural Language Processing (MEAP) this week in the Artificial Intelligence and Machine Learning forum!
  • Post Reply Bookmark Topic Watch Topic
  • New Topic
programming forums Java Mobile Certification Databases Caching Books Engineering Micro Controllers OS Languages Paradigms IDEs Build Tools Frameworks Application Servers Open Source This Site Careers Other all forums
this forum made possible by our volunteer staff, including ...
Marshals:
  • Campbell Ritchie
  • Tim Cooke
  • Paul Clapham
  • Devaka Cooray
  • Bear Bibeault
Sheriffs:
  • Junilu Lacar
  • Knute Snortum
  • Liutauras Vilda
Saloon Keepers:
  • Ron McLeod
  • Stephan van Hulst
  • Tim Moores
  • Tim Holloway
  • Piet Souris
Bartenders:
  • salvin francis
  • Carey Brown
  • Frits Walraven

Why is machine learning not used in the field of predicting stock price based on previous prices?

 
Ranch Hand
Posts: 1410
8
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Machine learning can be used to make predictions using different ways such as timeseries. Why is machine learning is not used to predict the stock price based on datasets of the prices of previous days? Is there any reason for this?
 
Marshal
Posts: 4856
317
IntelliJ IDE Python Java Linux
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
How do you know it isn't being used for this purpose?
 
Tim Cooke
Marshal
Posts: 4856
317
IntelliJ IDE Python Java Linux
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
I'll elaborate a little...

Automated trading is big business. It's also very secretive business because the algorithms form the primary asset for a trading firm. Therefore you will never really know what technology or techniques are being used.
 
lowercase baba
Posts: 12829
52
Chrome Java Linux
  • Likes 1
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
One thing to consider is that <today's price> is not simply a function of <previous prices>.  What's in a quarterly earnings report, a global epidemic, the announcement of new (or easing of) tariffs, OPEC announcing production cuts, Russia saying "I'm ignoring your production cuts", Saudi Arabia saying "Oh yeah? we'll produce even MORE to undercut your overproduction..." and on and on and on....

There are so many factors, often in a completely unrelated sector, that can affect the price of a stock.  I have no doubt people are trying, but I believe it is impossible to ever build an accurate model.
 
Monica Shiralkar
Ranch Hand
Posts: 1410
8
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator

the announcement of new (or easing of) tariffs, OPEC announcing production cuts, Russia saying "I'm ignoring your production cuts", Saudi Arabia saying "Oh yeah? we'll produce even MORE to undercut your overproduction...



Thanks.This accurately answers the question.  I think  quarterly earnings report we can include as some value in one of the parameters. And we can have a parameter as 0 or 1 based on whether global epidemic is absent or present but we cannot do anything about how to quantify the announcement or some leader's statement in terms of a parameter value. This part does not look possible. So algorithm will be very weak.
 
Saloon Keeper
Posts: 21963
149
Android Eclipse IDE Tomcat Server Redhat Java Linux
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator

fred rosenberger wrote:One thing to consider is that <today's price> is not simply a function of <previous prices>.



Or, as they say in the industry when they're trying to sell you something:


Past Performance is not an indicator of Future Results.



The process of attempting to predict stock price trends based on past history is called Technical Analysis and they have a whole raft of different terms for different graph profiles. But Technical Analysis is really only good for trying to respond to the trading patterns of other Technical Analysts. As Fred has said, a lot of non-technical factors also affect share prices.

In any event, automated short-term trading isn't something that's very useful for personal investors. The only way to make money doing that is to trade enough that small changes in share value add up to large totals. It's essentially a zero-sum game, however, allowing for overall market gains and losses.

It's pretty well accepted that if you want to make serious money you buy and hold for extended periods of time, which is what Warren Buffet does. Also know what you're buying, what your goals are, and where your purchases are likely to take you.

I paid off my house because I bought Cisco stock back when the Internet was beginning to expand. I didn't get it at the IPO or at some magic time where it was poised to take off, I bought it when I had money and it looked good. I sold it when I saw that the share growth pattern had levelled off. In the case of Red Hat I bought as soon after the IPO as I could and held it until IBM bought them out last year.

In neither case was I a financial genius. Rather, being someone who works in the industry, I looked for solid companies with a growth future and relied on my daily knowledge of what they were actually doing with customers - and planning to do - to determine whether to hold or sell. I didn't go with the front-line big names, although that does mean I missed out on Google and Amazon. I made the mistake of selling Apple - right before Steve Jobs returned. Ooops. But overall, it has done well. That kind of stuff is kind of hard to plug into a mathematical model, though.
 
Monica Shiralkar
Ranch Hand
Posts: 1410
8
  • Mark post as helpful
  • send pies
  • Quote
  • Report post to moderator
Thanks
 
Whose rules are you playing by? This tiny ad doesn't respect those rules:
Two software engineers solve most of the world's problems in one K&R sized book
https://coderanch.com/wiki/718759/books/Building-World-Backyard-Paul-Wheaton
    Bookmark Topic Watch Topic
  • New Topic