|From: jamie30000 (Rep: 0)||Date: 2016-10-30 12:19:20|
|Forum: Casual Lounge - Thread #673880287||Msg #713 - Part 1/2 (Rec: 0) |
First post so go easy please ;-)
I'm an experienced software developer with some time on my hands now and I'm looking to create a free system that will detect stock movements and then check pretty much every news source that can be checked online for news regarding this company. So that's blogs, social media, news websites etc. With the aim of getting this information to the user as quickly as possible.
At the moment I'm not sure whats the best way to detect that something might be effecting a stock, I was going to go with just the price of the stock but then I thought some news may be interpreted as good to some and bad to others so the price might not actually change much. Then I thought of checking the volume of the stock over a short interval. I saw some websites do offer this but it appears they don't show "out of hours" trading volume. Is that correct? If not can someone point me in the direction of one that does?
What stats would you guys recommend looking at to detect "interest" and where is the best place to source these stats/figures?
Any help at all would be greatly appreciated :-)
NOTE: This message was originally posted on 2016-10-30 08:15:33 by the jamie30000.
|From: kapivaktra (Rep: 0) reply to jamie30000||Date: 2017-07-19 19:38:11|
|Forum: Casual Lounge - Thread #673880287||Msg #745 - Part 2/2 (Rec: 0) |
|I know this was posted a while back but you did not seem to get any answers and that saddens me, but I suppose it may have just been in the wrong forum. I am new here and just joined today but wanted to get involved as quickly as possible. I am a technologist, so I am always intrigued by interesting questions regarding what tech can and can't do at the moment. |
Currently I would surmise that you would first define %boundaries based on price points. So a stock with a higher price point would need less of a % change to trigger your software (versus a stock with a lower price point) to "start digging." From here you would likely want to investigate volume or perhaps an overlay like the Ichimoku Cloud to give you direction and momentum of the movement. This is important because it lets you continually gather data for later use since you need to include machine learning in a big way.
Here is why you need to include machine learning. To aggregate keywords within a given article or blog post with direction and momentum (or movement) in a given stocks chart. This is important because as there is no way currently (that I know of) for a computer to tell you weather or not news about a company is bad or good much less how bad or how good (so the first step of linking the keywords is important before we can start assigning momentum to create a vector).
Likely this will take several years of trial and error and removing many false keywords like "the" and other items as well as ads and comments or directional indicators on comments such as thumbs up and thumbs down. Eventually, you may be in for a big problem though as many traders in the markets are using algorithms and software to decide if they trade or not on a given stock and the number is growing.
In short and to keep from being too long winded. I guess the gist would be; it just is not possible currently without massive funding and several dedicated team members to help you in this. As well, the technology we have at the moment just is not capable of computing if a given unit of news about company A will result in people divesting from company A, be it for fear of loosing funds, need of liquidity, wanting to distance themselves from what they may perceive as an ethical issue.
Generally speaking I believe it is currently done through assumptions and polling groups of investors and analysts if news about Company A was why they decided to sell, short, or change their given stance.
tl;dr without the ability to computationally link ethical values with language and to completely ignore the multiplicity of interpretation of language among human consciousness in favor of a binary interpretation model; it just is not going to happen. There is no way to link directly weather or not price point movement was precipitated by a unit of disseminated information, without direct communication with the individuals who moved the stock price point.
It does suck though that you came to ask one question and never came back when it was ignored...
NOTE: This message was originally posted on 2017-07-19 19:33:35 by the kapivaktra.
| Reply to jamie30000 - Msg #713 - 10/30/2016 12:19|
TheLion.com | About Us | Agreement & Disclaimer | Privacy | Twitter
© 1999- TheLion.com, Inc.