Is this some kind of joke? How is amazing that the prediction is just following the price and thus completely useless?
Also most of the indicators you are suggesting do not really affect the price. The contrary is usually true instead: The price does affect those "indicators". Which basically means they were not leading but lagging indicators in first instance. Ie: Google Trends, etc....
Topic: Using Deep Learning For Predicting Price. [NEED HELP] (Read 309 times)
Upon checking your graph, comparing the two data sets, the prediction is delayed for a day. It wouldn’t be much help if it had already happened before predicting it, right? Have you used the actual forecast and made some trades with it? I suggest that with your particular data, correlate the date and the one news about cryptocurrencies that have been published on that same day, and probably you could have an idea with the reports that have affected bitcoin so much. At least after that, you would have an idea if you have a specific news article that could correlate with your data, then you could predict it more accurately.
That is what I was thinking as well, the price and the prediction looks very close but we know that in any market being close is not enough, the OP doesn't really give any indication that he has a way to take advantage of those predictions, I remember that I did something similar some time ago and while the predicted price was not that far away I was completely incapable of creating a strategy around that and be profitable so I dropped that idea really quickly.
Probably try multiple deep learning analysis then compile them into one system that analyzes the entire concept of the said combination? There's a lot to take into when using Deep Learning and having data as the news received since a person could have different opinions regarding a news depending on their country of growth. Taking the number of countries in the world, plus the amount of news circulated everyday, that would take a enormous amount of processing data that would probably put a strain no matter how fast or big a system is. Probably be better if categorized based on country and type of news, whether it be a positive, negative or something of the sort.
Just throw anything and everything tbh. Deep Learning is only useful and accurate IF and only IF it has access to multitudes of data in terms of different perspectives and analysis. It requires a huge amount of data and space to store everything, but It could potentially be quite accurate in terms of prediction, just that if it actually failed to take in one factor, for example, a case of intervention by a single man whom seriously affected the market price, the prediction made using deep learning would fail, and quite remarkably with that.
As for the specifics of news, tweets, etc., its honestly difficult to take them all into account. I mean, everyone views different news, everyone views different sources, and if you take into account the number of people who view a certain source, that all together makes gathering info already difficult. Probably have something similar to a watch dog that removes FUD to FOMO to relevant news.
As for the specifics of news, tweets, etc., its honestly difficult to take them all into account. I mean, everyone views different news, everyone views different sources, and if you take into account the number of people who view a certain source, that all together makes gathering info already difficult. Probably have something similar to a watch dog that removes FUD to FOMO to relevant news.
Source: https://www.kaggle.com/c/two-sigma-financial-news
My question may sound stupid, but what news, posts, tweets, reports mostly affect at BTC price? Because the problem right now in training data and in weight for news.
My question may sound stupid, but what news, posts, tweets, reports mostly affect at BTC price? Because the problem right now in training data and in weight for news.
I looked at a couple of the example submissions and my initial reaction is: isn't your self-learning network supposed to find correlations to determine that, given a multitude of data? Shouldn't you just be throwing everything at it, and seeing what sticks? Anything related to economy, markets, technology, the cryptocurrency and token service industries, for starters. Widely distributed mediums are obviously more interesting than items nobody reads.
The legend for news data suggests the model could be fed seemingly irrelevant news that doesn't mention the asset in question. The model should, over time, de-emphasize irrelevant news that has no effect and emphasize news that does has some effect. The model shouldn't depend on humans determining what's important and what's not. That defeats the purpose.
While the historic data is used by many traders to predict crypto prices, there are also some other data that is used by traders which help in making predictions.
Many people use google search data to make predictions based on it and gain profits.
Many people use google search data to make predictions based on it and gain profits.
Quote
My question may sound stupid, but what news, posts, tweets, reports mostly affect at BTC price? Because the problem right now in training data and in weight for news.
Posts/Tweets from Celebrities and big players/Entrepreneurs also make an impact on crypto prices many times. 
There is a twitter account that shares all the whale movements, anything above 1 thousand bitcoins moving is a tweet by them it is automatic I think and that looks like the most influential bitcoin prediction there is. https://twitter.com/whale_alert, this is their address and if you could add that and combine the fact that whenever price moved up or down and whenever there was a tweet of movement, if you can find some correlation between those two you may actually end up training a program to predict what will happen as soon as there is a movement and act accordingly.
Since, humans can't watch a twitter account 7/24 and since we can't move as fast as a trading bot, when you build a trading bot that takes a look 7/24 to that twitter and buys/sells accordingly instantly you can actually beat the whales on their own game.
Since, humans can't watch a twitter account 7/24 and since we can't move as fast as a trading bot, when you build a trading bot that takes a look 7/24 to that twitter and buys/sells accordingly instantly you can actually beat the whales on their own game.
You might want to check out the Cindicator project. They are using "Wisdow of the crowd" and machine learning/artificial intelligence to develop trading algorithms and strategies.
Almost every indicator out there has a correlation with the market. After all EMAs, RSI, Stochastic, etc. are all based on the open and close, volume and time. The only way to "train" a machine/algorithm is by intensive tests.
Almost every indicator out there has a correlation with the market. After all EMAs, RSI, Stochastic, etc. are all based on the open and close, volume and time. The only way to "train" a machine/algorithm is by intensive tests.
Hello everyone,
Recently, I found some articles where people tried to predict prices using Deep Learning, only using data (Close Price, High Price, Low Price, Open Price, Volume) and the result was amazing :
Source: https://towardsdatascience.com/cryptocurrency-price-prediction-using-deep-learning-70cfca50dd3a
But if also add technical indicators to training data sets and the question is what indicators or sets of indicators should I use? (except RSI and MACD).
Also a few months ago, Kaggle (a platform that hosting machine learning competitions), had competition started by Two Sigma: "Two Sigma: Using News to Predict Stock Movements" that ended also "well".
Two Sigma provided 2 data sets:
So, all competitors have trading data for their neural networks.
Source: https://www.kaggle.com/c/two-sigma-financial-news
My question may sound stupid, but what news, posts, tweets, reports mostly affect at BTC price? Because the problem right now in training data and in weight for news.
I believe, that "symbiosis" of Neural Networks trained using data from past prices moves, technical indicators, and news, will create a powerful mechanism for profitable trading.
And it would be also cool to find enthusiasts, that think the same.
Most Deep Learning models now use old data from price charts to predict the market. However, we all see how news affect on traders/ users, especially newbie. It would be great if we can use these data, it would improve the model accuracy. Recently, I found some articles where people tried to predict prices using Deep Learning, only using data (Close Price, High Price, Low Price, Open Price, Volume) and the result was amazing :
Source: https://towardsdatascience.com/cryptocurrency-price-prediction-using-deep-learning-70cfca50dd3a
But if also add technical indicators to training data sets and the question is what indicators or sets of indicators should I use? (except RSI and MACD).
Also a few months ago, Kaggle (a platform that hosting machine learning competitions), had competition started by Two Sigma: "Two Sigma: Using News to Predict Stock Movements" that ended also "well".
Two Sigma provided 2 data sets:
- Market data provided by Intrinio.
- News data provided by Thomson Reuters.
So, all competitors have trading data for their neural networks.
Source: https://www.kaggle.com/c/two-sigma-financial-news
My question may sound stupid, but what news, posts, tweets, reports mostly affect at BTC price? Because the problem right now in training data and in weight for news.
I believe, that "symbiosis" of Neural Networks trained using data from past prices moves, technical indicators, and news, will create a powerful mechanism for profitable trading.
And it would be also cool to find enthusiasts, that think the same.
there have always been some people who used past data to predict the market, by using the data Close Prices, High Prices, Low Prices, Open Prices, Volume it can see the current trends, but there are no predictions that are truly 100% accurate, technical analysis sometimes also often deceives.
Past data was frustrating, better look upon the latest price progress which is nearly possible instead of seeking complicated analysis. You can do it successfully by your own as long as you're doing the best that you can. Of course patience is very important, and by developing it at your own perspectives you must rely with the live market updates.
checking graph data many its lagging with the actual data and many moving averages lower value of 20 also gives similar movement as moving averages also uses data of open close low and high , all this indicators works when tread and volume is good , ca you update with some more graph of when volume is low
I think we cannot use the term deep learning because all we know that even what research and study that we would do from the past prices in the end it will fall in guessing. No one can predict about the prices because if someone can predict crypto industry is already dead and the one who can predict is the most richest man in the world.
Therefore the prediction is only for a temporary guess while the price cannot really be guessed by any expert, sometimes I am also surprised why many people believe in the prediction even though it is not necessarily true in their guesses and it is only their speculation to motivate themselves to be more just positive.
The price of bitcoin no one can guess anyone because the price fluctuations can not be relied on.
Upon checking your graph, comparing the two data sets, the prediction is delayed for a day. It wouldn’t be much help if it had already happened before predicting it, right? Have you used the actual forecast and made some trades with it? I suggest that with your particular data, correlate the date and the one news about cryptocurrencies that have been published on that same day, and probably you could have an idea with the reports that have affected bitcoin so much. At least after that, you would have an idea if you have a specific news article that could correlate with your data, then you could predict it more accurately.
I think we cannot use the term deep learning because all we know that even what research and study that we would do from the past prices in the end it will fall in guessing. No one can predict about the prices because if someone can predict crypto industry is already dead and the one who can predict is the most richest man in the world.
Pretty much anything positive on the news about Bitcoin will affect the price. For instance, I'm sure very soon we'll start seeing news articles about how Bitcoin is up 100% this last year. That will in turn bring more people in that may have never even heard of Bitcoin. Now, the opposite is true whenever something bad hits the news. For instance, Mt. Gox getting hacked. That would make people worried about investing in Bitcoin because they may think that their funds could get stolen and usually brings a dip in price along with it.
there have always been some people who used past data to predict the market, by using the data Close Prices, High Prices, Low Prices, Open Prices, Volume it can see the current trends, but there are no predictions that are truly 100% accurate, technical analysis sometimes also often deceives.
Hello everyone,
Recently, I found some articles where people tried to predict prices using Deep Learning, only using data (Close Price, High Price, Low Price, Open Price, Volume) and the result was amazing :
Source: https://towardsdatascience.com/cryptocurrency-price-prediction-using-deep-learning-70cfca50dd3a
But if also add technical indicators to training data sets and the question is what indicators or sets of indicators should I use? (except RSI and MACD).
Also a few months ago, Kaggle (a platform that hosting machine learning competitions), had competition started by Two Sigma: "Two Sigma: Using News to Predict Stock Movements" that ended also "well".
Two Sigma provided 2 data sets:
So, all competitors have trading data for their neural networks.
Source: https://www.kaggle.com/c/two-sigma-financial-news
My question may sound stupid, but what news, posts, tweets, reports mostly affect at BTC price? Because the problem right now in training data and in weight for news.
I believe, that "symbiosis" of Neural Networks trained using data from past prices moves, technical indicators, and news, will create a powerful mechanism for profitable trading.
And it would be also cool to find enthusiasts, that think the same.
Recently, I found some articles where people tried to predict prices using Deep Learning, only using data (Close Price, High Price, Low Price, Open Price, Volume) and the result was amazing :
Source: https://towardsdatascience.com/cryptocurrency-price-prediction-using-deep-learning-70cfca50dd3a
But if also add technical indicators to training data sets and the question is what indicators or sets of indicators should I use? (except RSI and MACD).
Also a few months ago, Kaggle (a platform that hosting machine learning competitions), had competition started by Two Sigma: "Two Sigma: Using News to Predict Stock Movements" that ended also "well".
Two Sigma provided 2 data sets:
- Market data provided by Intrinio.
- News data provided by Thomson Reuters.
So, all competitors have trading data for their neural networks.
Source: https://www.kaggle.com/c/two-sigma-financial-news
My question may sound stupid, but what news, posts, tweets, reports mostly affect at BTC price? Because the problem right now in training data and in weight for news.
I believe, that "symbiosis" of Neural Networks trained using data from past prices moves, technical indicators, and news, will create a powerful mechanism for profitable trading.
And it would be also cool to find enthusiasts, that think the same.
Jump to: