So I built a bot

Far better than anything I could do bud, alphabetize them and it will be better

You can use AI features as plugins. For example Watson. They can they return a sentiment value.

Thanks for the idea. Imma do the same as op except with sentiment analysis and opensource it. Good excuse for me to brush up on my golang

>How are you scraping the data?
>I can tell you're not a programmer.
kek

Sure man keep /biz posted

Go for it man

It's brilliant honestly.

I'll provide capital for this idea if someone gets a good business plan going.

I could easily get a grant from the government for this and I don't mind using some of my savings to get something like this on the market

don't break your head on sentiment analysis, there are API for that.

Check this one out: ibm.com/watson/developercloud/alchemy-language.html

Watson Alchemy

youtube.com/watch?v=OG3WMaWwJig

you makeh me angery

how am i meant to buy lambos if mofos keep opensourcing shit.

anyway i wish you the best good luck

so i could copy and pasta "mooncoin" 100 times in 1 post and it'll come up as 100 hits?

or is it limited to 1 per post?

You're creating an oracle right? have you read up on aeternity.com and how they're going about it? They're using golang for that too I think

You should test it out.

I'll help:
mooncoin mooncoin mooncoin mooncoin mooncoin mooncoin mooncoin mooncoin

digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte digibyte

I guess I could limit it to one coin extraction per post.. thanks for the idea man.. appreciated

thanks for link, howeve rI believe sentiment analysis would just confused my algorithm as shillers tend to be liars D:

does it track memes?

ok testing now

hmmm thats weird they havent appeared, maybe oh shit yeah Im not doing a lowercase check

SO we could make a thread and in every post spew out every coin in existence to screw with it?

This is a gud thread. Keep going user! Bring Veeky Forums to the moon!

fuck I dont know why digibyte isnt being picked up

it looks like mooncoin is, but not 8 times... wtf.... god damn bugs

Also can you scrape associated keywords like good,great,shitcoin,project,active,developers,dev,competition,win,news etc etc and pin them to the coins?

mfw ive been working on this as well

yeah thats easy enough, but would that help? wouldnt that shit just confuse hte algorithm because of lies?

Basically I have a couple other sites Im going to cross reference my algo with to create a holy grail of an app ( im an optimist) , so if you guys have any ideas on what else I could cross reference with so we dont just get an algo poisoned by bag holding shills, then that would be great.

im sorry i have no idea why digibyte isnt showing =/

it doesnt do image analysis (yet... i dont think I will do thibit though) but lol pepe image recognition sounds funny. would it be good or bad to have a pepe shilled coin?

These screengrabs are only of the program running righ tnow, on the front page ( not all the pages ) maybe the thread gets pushed down and digibyte isnt detected?

Are you scraping the data directly from Veeky Forums?

I can't imagine it would be difficult to create a live website that updates instantly from the past 24 hours

Do case insensitive searches, and also link abbreviations and whole words together, so that "eth" and "ethereum" are not 2 different entries.

subscribe to the mailing list dude, it seems to be getting popular. If you want to email me, just wait for my first mail out and reply to th email

ok I ran a bigger build, for a few more biz pages, the results seem to change more than I expected

I could do that but Im not sure whether that is a priority at this stage, ill add it to the todo list

Dank,
What will you include in the 'final' build; do you think you'll be able to do the entirety of Veeky Forums?
Also how many people on the mailing list now if you don't mind saying?

lol, I can code this in java if you guys want and put it up on github for free, with BTC donation address of course :>

Anyone down? I can get it done in about an hour or two

It can do the entirety of biz, its just like thousand posts so it takes about 10 minutes to run it. Ill start doing full runs with the mailouts once a day to the subscribers first.

Subscribers are just those who have participated in the thread I reckon, not like lots.

Kekspeed to you user

ok I scanned the whole board, here are the results.

Good night everyone , hopefully Ill get the mailout working in the morning.

Go away /g/

I actually made the same thing and linked it to info from coinmarketcap

some nice ideas in this thread for additional info

java does everything and is robust. Also why make it a mail list and not a public webservice?

lol, have you never heard of clif high?

He already has a product like this but scans sources way beyond Veeky Forums. He makes s killing on this scam too, some of his predictions are pretty damn accurate though, an exact halving may 22 - may 28th for ETH and that happened.

ignore that guy gj its cool

Great idea OP. I'm going to copy this idea and make a much better version.

There is already a coin for that. cryptoping.tech/

did you buy his report?

His site looks outdated and sketchy as fuck.

I should do the same thing and wrap it in a more legit looking package

SHUT IT DOWN

fuck no, it's not hard to get a fucking pdf document online

how can i find it?

I'm actually working on an AI as we speak that scrapes Veeky Forums and other forums, conducts sentimental analysis on shilled coins and does forecasting (monte-carlo) based on the sentiment

What websites does he scan for? I'm going to guess reddit right

The entire internet

if(op==faggot) {
killyourself;
}

>someone stole my idea

Fuck

What are you using for the AI

I just got a good eth domain to go with it for donations too

post it

Why don't you add an infinite loop in there.

To be sure he stays dead.

>what else I could cross reference with
can you plugin to twitter trends data?

Pastebin the code please OP.

biztrack.eth

I don't have any btc senpai but I will in the future.

From my limited programing experience, I think I can keep him dead.

while (op == faggot){
killyourself;
opDeath++;
ConsoleWriteline("OP died: " + opDeath + " times");
}

lol

let me help

while (OP.ID == 'V1uzlO8z' || user!=subscribed && OP.ID==''E8S1iixX"){
OP.killyourself();
OP.Death++;
ConsoleWriteline("OP died: " + OP.Death + " times");
}

...

>OP.ID == 'V1uzlO8z' || user!=subscribed && OP.ID==''E8S1iixX"

are you retarded or something?

Holy shit OP you're going to make bank off this, all these gullible retards giving you their email addresses. You're going to have some premium mailing lists to sell to advertisers, have you thought about making a daily text messaging service too?

...

you do realize that OP.ID == 'V1uzlO8z' && OP.ID==''E8S1iixX" can never be true right, so the check for OP.ID == 'V1uzlO8z' is completely unnecessary

btw dude I am happy to collab, reply to my mail out.

people tend to sign up with throwaways. literally the only way ill make money it seems is if I get my follow up algorithm to actually work . key is always in the details and the implementation.

>while (OP.ID == 'V1uzlO8z' || user!=subscribed && OP.ID==''E8S1iixX"){
>OP.killyourself();
>OP.Death++;
>ConsoleWriteline("OP died: " + OP.Death + " times");
>}
ah shit, sorry compiler man. Here we go:

while (OP.ID == 'V1uzlO8z' || (user!=subscribed && OP.ID==''E8S1iixX") ){
OP.killyourself();
OP.Death++;
ConsoleWriteline("OP died: " + OP.Death + " times");

}

Do some UX design so I don't gauge my fucking eyes out. It isn't 1980 bitch.

Using TensorFlow to train the model. Going to scrap Veeky Forums and other sources about 2 years back and correlate it with coinmarketcap data. Then train the model on the combined dataset. I'll then incoporate this into a real-time service to make predictions based off of forum sentiment - apparently 80% of market movement is a result of forum activity

get out of this thread you literal fucking autists.

blog.stageslearning.com/blog/a-new-education-teaching-coding-to-students-with-autism

is there an archive of coinmarketcap data?

what kind of neural net do you use?

Hey mate, I'm going to be posting my source in this thread in a few hours. Wanna collab?

I'm hoping I can build a scraping tool which will extract data from the coinmarketcap charts (e.g. 7th November 2015 it will scrap all coins on market for current cap, volume, price etc.).

Will also be using a recurrent neural net. If I can get it to work then I'll be adding in technical indicators (MACD, RSI etc.) into the mix too and see what the results are like

Where did you study to learn all this?

Neat, sounds exactly like the approach I was gonna use, though I am using dl4j on top of spark instead of tensorflow

You can find all the information online really, there's many machien learning tutorials out there now.

I don't quite understand.

What is a "moon coin"?

Is it for buying from bottom baring businesses?

Looks good though anyway well done!

I see the thread has now filled with all sorts of weird CS shills loudly proclaiming how they are going to hook up regression modelling with tensorflow and create a portal into a new Biz dimension.

Dont forget to add a coffee cup holder guys.

>"apparently 80% of market movement is a result of forum activity"

is this true? anyone have a source for this?

gave me a keke tbqh lad

reddit, just google it

Bum coins

ibm.com/watson/developercloud/doc/tone-analyzer/index.html

If this takes off, you should definitely share the wealth of knowledge here on Veeky Forums. If enough Veeky Forumsrealites get rich we can all share the wealth.

Were you planning on sharing the info through an email subscription like OP or were you just going to do it for personal data analysis?

If you had the possibility to predict the market, why would you think he would share it?

guessing that's a joke. i already had googled it and found nothing so im presuming its not true

Just to be helpful/kind enough to share I guess. He could also charge a subscription for the info but that's assuming he would go full Schlomo Shekelstein on everyone.

>Just to be helpful/kind enough to share I guess
lmao fuck off

literally took me 10 seconds

reddit.com/r/Bitcoin/comments/6cfwir

You realize YOU'RE being the schlomo for wanting free shit right

You're not wrong.

maybe later, this thread has potential.

Already started my own implementation of this idea.

Got all the pulling data from Veeky Forums down and some basic sentiment analysis going. Needs many hours of work before I'd feel as though I'd have a product I could charge money for, so I may share what I have when im a bit further along and have a basic PoC.

ah, the google fu is strong with this one :) thank you, kind sir, and enjoy the reddit gold!

anyway, unfortunately (or maybe fortunately considering we're on Veeky Forums), this article says that it's not that 80% of price movement is influenced by forum posts, it's that's some Korean software was just able to correctly predict the price movement 80% of the time based on forum posts on bitcointalk.org.

but the question we're all wondering: can it predict bitbean's?

OP, I've thought about doing this a few times. It's a pretty good data analysis tool to see what's being shilled, and given enough time you might be able to calculate entry / exit points. Behold, the shill oscillator.

Mailing list is kind of 'meh' though, this needs pushed to a web front-end of some kind. Like other anons said, It's 2017 bro.

Clif's methods are a little different, supposedly he scrapes the net for repeated terms and assigns them an 'emotional sum' based on the terminology around it, or something like that. His interpretations absolutely suck however, while for crypto he's been doing OK, many of his calls have been total garbage and are likely clouded by personal bias.

IIRC he's saying something like $13k BTC in Q1 2018 (But of course, if you want to know more, now you need to buy his $100 reports.).

I think the tools he is using should be explored more rigorously (Like, actual academic papers, etc.); but at the level he's using them it's only about as good as a horoscope.

not to mention he is, by his own admission, poor.

How do you plan to scan forums that need a login?