Topic: Research Price Bitcoin Project (Read 238 times)

I agree with you, the amount of knowledge needed for this project is impressive: statistics, mathematics, machine learning studies, market knowledge and not to forget programming. For the prediction algorithm I worked for 8 months, and only for that. Right now I'm making the changes after getting a lot of data in the previous months

interesting, then there will be the way even in that "world" to invent and do something new .... Today I developed this project, maybe in that scenario I'll do something else!

new communist movements and coins of style area with new banking cartels arising

Probable
So, the end will be a confusing world of cryptocurrencie, and so, what will we do?

In twitter the chart is possible only within an image...I will probably link the tweets to a web site with the information you suggested. Thank you!

i doubt that bitcoin will stay an important cryptocurrency for long, people arent supporting it anymore just look at the second and third bitcoin fork (bitcoin gold -96% bitcoin cash -80%)

they are dying and bitcoin will follow them,

what stays after that will be a babylon like world of confusing cryptocurrencies

It is better to output these data in a statistical form or to update the data in the form of charts rather than digitally displayed. [make a chart like CoinDesk]

 - Try to make the error between the expected and actual values given the correct results in future predictions.
 - Try to be an analysis based on the news and then improve the algorithm to ascertain the amount of impact of any news on the price and thus become more credible prediction.
 - Provide pricing charts for historical data, news, and trading platforms. Any abnormal difference would be a manipulation of the price.

Good work looking forward to upcoming updates

Well, a project that I will follow ...but remembre my project for now is experimental

Of course, the algo KNN is used together with sentiment analysis (for now only twitter).

As far as am concern, Bitcoin future rest heavily on human sentiments as there is o clear market analysis at present. Am struggling to understand the basis of this project. If there is any tool that performs sentiment analysis, then I see hope in it else it will amount to futile effort.

I have used some genetic algorithms to search for patterns, but now I'm trying to extract a good result (with extreme difficulty) from the prediction algorithms.

The algorithm analyzes historical bitcoin data and use K nearest neighbours algorithm to make a prediction (very simple...for now is experimental project). I use the Bitcoin historical API powered by CoinDesk to collect the data. In the future, I would to add a sentiment analysis to this value.

interesting arbitrage seems to have lost its appealing i can remember doing 10% profit with that per day

I've done both - I've built an arbitrage bot and an AI that finds trading patterns (genetic approach).
The arbitrage bot isn't efficient because the markets are moving so fast that you don't get the prices that you want - even if your sever is directly connected to the exchange server. It doesn't matter how fast your algorithm is - it's physically impossible (talking about bitfinex). You can maybe earn something on slower markets but it's less lucrative and it takes a lot of work to do and event after that it isn't guarateed to work.
It can also be that you have calculated that you will earn a lot of money but after a few months, when the project is finished it's 1/10 of that because numbers like trading volume have changed significantly. What, maybe, could work are HFT strategies implementing a mix of insurance Mathematics and Ergodic theory which is a bit like trying to beat roulette at the casino but this takes a LOT of work and time.
Cross exchange arbitrage can end up being a huge problem because sometimes coins are locked for deposit and withdrawal and then people will pump them usually. Unexpected exchange behaviour in general is a problem. And I haven't even started to talk about all the numbers like fake volume that a lot of exchanges are displaying (like 95% fake volume).

The AI is funny - the more complex your trading strategy becomes, based on the data you feed the bot with, the less it will work in the future. It's because you have the too opposites "General Rule" vs "Specific rule". The more specific a rule is the less it will work in general and the more general a rule is the less it will work in a specific case. So what the Bot wants to find is a (relativly) simple, general rule. A rule like that does not exist. By saying "does not exist" I mean that there is no simple, general rule that would have made you more money than just buying and holding bitcoin (if you analyze all bitcoin trading data on Bitstamp from 2015 until now)

i think coding an interexchange arbitrage bot is more secure, and more successful than trying to prognose bitcoin

but there are many doing that already

Yes, in principle I agree with you, but I would like to try. The project as I pointed out is experimental. It's in an embryonic phase, and right now I'm analyzing the results. The goal is to understand how accurate the proposed value can be.

Artificial intelligence algorithms? interesting.
have checked the twitter cryptosignal and it's quite confusing, how does it always changing to predict bitcoin price within a day?
Ahh, got it. seems the prediction always recalculate after 2 hours, assume it pretty accurate then.
what kind of artificial algorithms did you write to ensure it will calculate and analyze correctly?
and based on what kind of information? whether any good news or bad news will make any difference?
or maybe trading volume on some big exchanges will affect the calculation?
One more thing, we have dozens of exchanges out there, thus the price of bitcoin slightly different, which exchange or platform as your reference?
Bitcoin price on coinmarketcap right now is $6409, on preev is $6355, on coinbase is 6352, OKCoin is $6455, meanwhile on local exchange in my country is $6545.

the market isnt systematic or periodic, like natural life and death cycles....., bird winter refuge routes, or salmon breeding return wanderings

it cant be described with algortihms...

there are trends, fashions that are temporarily.

In fact, I have highlighted how the project is an experiment. However, machine learning algorithms can help in short-term trading decisions due to the ability to analyze large amounts of data in fractional seconds. In the long term it is obvious that this can not happen.

these are assumption based on

1. limited information

2. based on a made up own forumla

both are factors that make it invalid.

you cant prognose bitcoins worldwide acceptance in the world market
as you neither can do that for the us dollar

regards