Julia for Cryptocurrencies — $EOS elasticity in a linear model
I´m new in Julia. So, let me know what I must try in Jupyter Notebook. Please, comment this text. For a while, keep on moving the same spirit I have in Python: coding to track cryptocurrencies. If you code in Python, try Julia too. You’ll love it… Some days ago, I did it in JuliaBox (log in for a free subscription model, no installations, few compute but with a very nice mindset) and found an interesting fact about $EOS. I have 48.45% of my cryptofolio in $EOS. That’s why I study it coding. First, I must say it’s a beautiful project that has chosen a brave way. I like to think differently in a competition ecosystem too. Well, once in Jupyter Notebook by JuliaBox (New — Notebook), let’s import our useful packages and load the data in Julia. We have to do something like:
I chose the way “using Queryverse” to load $EOS data in a csv format from CoinGecko.com and transform it into a DataFrame (I’ve just commented a way that works too). You can also try “using DataFrames, Requests, IterableTables, GLM, StatPlots”. No problem. We need GLM to create a linear model (regression) and StatPlots to show it. Queryverse pkg has a plot tool too so StatPlots is redundant. As I said before, I’m new here. Julia versatility is unbelievable. I know there can be a best way to run it. By any chance, I don’t have any idea what the WARNING tells me. Just ignored it. You can see much more here.
Run a cell “describe(df)” to see Summary Stats of all columns. In Julia, we can apply a function to a group of columns at the same time by using “colwise( )”. For example, I run “colwise(std, df[[:2, :3, :4]])” to know the standard deviation of ‘price’, ‘market_cap’ and ‘total_volume’ columns. See an introduction to Julia/DataFrames here. If you can buy more compute, try “df |> Voyager ( )” to print out all statplots of your data (only if you load datasets I guess). My pleasant surprise started when I ran “cor(df[:2], df[:3])” to find the correlation coef between ‘price’ and ‘market_cap’. I found 0.9749197610686224, a very strong positive correlation I’d never seen before in cryptocurrencies I used to trade. So, I wanted to see it in a linear regression model.
I guess so funny when geek experts talk about Machine Learning with so much Math and Stats as ML could give you all precise answers you dream. Imagine a sad emoji here. No, it can’t. But we find out the best financial trends when we code ML models for cryptocurrencies. Indeed, I started to code in Python language to track those trends. Now, here comes Julia. Linear Model — done. Summary, estimated intercept and the estimated slope — seen. I created a linear function f(x) to plot it. Voir le jour dans cryptos…
I’d like to see any counter evidence here. Please, comment it. Otherwise, I can see a lovely $EOS price elasticity. It’s not a good model for predictions because price data moves away UP from the line. That’s it. The price goes up so hard when market capitalization increases. Today (June, 22 — 2018), $EOS market cap is almost US$8BI. Bitcoin, US$106BI. $EOS price is US$8.90 but what I’m trying to say, despite of excitement by Julia, it can hit US$20 easily in a bullish market. Am I so wrong?