Predicting Cryptocurrency Prices with Fibonacci Retracement and RNN
The frequency of posts has declined on this site over the past couple of weeks. The explanation is simple – I have become obsessed with my latest project:
Predicting Cryptocurrency prices using Fibonacci Retracement and Recurrent neural networks (RNN).
First – this post isn’t a tutorial or a how to predict the cryptocurrency market(s) post. The lovely Teresa Gail will be furious if I give away too much…
Second, this is in no way financial advice or a get quick rich scheme.
I write code – I am not any sort of financial advisor. Additionally the boom times for crypto is over; (hopefully) the market has matured so if you are looking to “Get Rich Quick with Bitcoin” this is not the article for you.
What it is: it’s more of a what I’ve been up to and am excited about post.
That being said, if you are tech savvy, have a basic understanding of the crypto-market, and are interested in learning more about my work with Crypto price prediction and AI – get in touch.
My Focus on AltCoins
For lack of a better definition you can think of alt coins as any blockchain based digital currency that is not Bitcoin. There are hundreds of other cryptocurrencies out there – known as “altcoins,” or alternatives to bitcoin.
Bitcoin is expensive. I will not be experimenting with Bitcoin strategies any time soon. AltCoins however have a much lower barrier to entry, for example, right now Ripple (XRP) is $0.43 per coin while Bitcoin (BTC) is $7,960.00
Some alternative blockchains even have “testnets” that allow you to experiment with pretend crypto.
If you need help getting started with Cryptocurrency feel free to contact me – I can provide consulting for a very reasonable fee.
The Altcoin Market as a Whole
One of the more interesting things that one can do with time-series data is predict estimated trend lines. Trend estimation is an effective way of creating models of time series data to analyze, and in my case feed into AI.
In my project the AI takes the shape of a custom Recurrent Neural Network.
A thought-provoking way to look at cryptocurrency is like it is a stock market – where each coin is a different stock. The code that produced the chart above was very helpful in creating trend lines for a subset of the alt-coin market.
Predicting Ethereum Prices
I could write an entire blog post on why I am interested in / focused on Ethereum. In fact, look for one within the week… For now, lets just assume I have reasons for choosing ETH as the focus of my research.
My latest version of this project uses a Recurrent neural network tuned specifically to predict Ethereum (ETH) prices based on a history of time series data and custom trend estimations.
My results have been encouraging indeed. In addition to proper ETH I have been using the Rinkby Testnet as well as Ropsten to try different strategies based on my RNN output.
That’s my allotted 500 words, which means it is back to my code…