How to implement this Bayesian Regression Algorithm?

Questions
, , , ,
1

I want to implement a Bayesian Regression Algorithm (see the link here: http://arxiv.org/pdf/1410.1231.pdf ) in Mathematica to predict Bitcoin prices (BTC/USD pair). I want to implement it on historical BTC/USD (Bitcoin/US Dollar) price data sourced from Quandl.com in Excel format. I have attached here both the algorithm document (PDF) and curated BTC/USD 1012- 2014 price data(Excel). Plz help!

2

This is a full project and a bit too much to cover in a single thread on a forum. Let’s begin by breaking your task up.

  1. Can you import the data from Quandl? Quandl has documentation on how to use their API. You can use the Quandl API with either the Import function or the URLFetch function. What is the URL for the query you want to make from the Quandl API?

  2. Let’s first try some simple predictive modeling on the time series before trying to implement an algorithm from a paper. Can you use TimeSeriesForecast on your data?

3

Hi Sean,

Thanks for your response. I want to fetch the following URL: https://www.quandl.com/data/BITSTAMP/USD-USD-BITCOIN-Exchange-Rate
Alternatively I want to import near real-time BTC/USD market price with 15minutes delay from: Blockchain Developer APIs

{
  "USD" : {"15m" : 478.68, "last" : 478.68, "buy" : 478.55, "sell" : 478.68,  "symbol" : "$"},
  "JPY" : {"15m" : 51033.99, "last" : 51033.99, "buy" : 51020.13, "sell" : 51033.99,  "symbol" : "¥"},
  "CNY" : {"15m" : 2937.05, "last" : 2937.05, "buy" : 2936.25, "sell" : 2937.05,  "symbol" : "¥"},
  "SGD" : {"15m" : 605.39, "last" : 605.39, "buy" : 605.22, "sell" : 605.39,  "symbol" : "$"},
  "HKD" : {"15m" : 3709.91, "last" : 3709.91, "buy" : 3708.9, "sell" : 3709.91,  "symbol" : "$"},
  "CAD" : {"15m" : 526.72, "last" : 526.72, "buy" : 526.58, "sell" : 526.72,  "symbol" : "$"},
  "NZD" : {"15m" : 582.26, "last" : 582.26, "buy" : 582.1, "sell" : 582.26,  "symbol" : "$"},
  "AUD" : {"15m" : 524.61, "last" : 524.61, "buy" : 524.46, "sell" : 524.61,  "symbol" : "$"},
  "CLP" : {"15m" : 283014.81, "last" : 283014.81, "buy" : 282937.95, "sell" : 283014.81,  "symbol" : "$"},
  "GBP" : {"15m" : 297.4, "last" : 297.4, "buy" : 297.32, "sell" : 297.4,  "symbol" : "£"},
  "DKK" : {"15m" : 2756.84, "last" : 2756.84, "buy" : 2756.09, "sell" : 2756.84,  "symbol" : "kr"},
  "SEK" : {"15m" : 3403.41, "last" : 3403.41, "buy" : 3402.49, "sell" : 3403.41,  "symbol" : "kr"},
  "ISK" : {"15m" : 56797.78, "last" : 56797.78, "buy" : 56782.35, "sell" : 56797.78,  "symbol" : "kr"},
  "CHF" : {"15m" : 447.19, "last" : 447.19, "buy" : 447.07, "sell" : 447.19,  "symbol" : "CHF"},
  "BRL" : {"15m" : 1093.06, "last" : 1093.06, "buy" : 1092.77, "sell" : 1093.06,  "symbol" : "R$"},
  "EUR" : {"15m" : 370.13, "last" : 370.13, "buy" : 370.03, "sell" : 370.13,  "symbol" : "€"},
  "RUB" : {"15m" : 17806.28, "last" : 17806.28, "buy" : 17801.44, "sell" : 17806.28,  "symbol" : "RUB"},
  "PLN" : {"15m" : 1557.38, "last" : 1557.38, "buy" : 1556.96, "sell" : 1557.38,  "symbol" : "z?"},
  "THB" : {"15m" : 15398.04, "last" : 15398.04, "buy" : 15393.86, "sell" : 15398.04,  "symbol" : "?"},
  "KRW" : {"15m" : 494436.55, "last" : 494436.55, "buy" : 494302.27, "sell" : 494436.55,  "symbol" : "?"},
  "TWD" : {"15m" : 14340.68, "last" : 14340.68, "buy" : 14336.79, "sell" : 14340.68,  "symbol" : "NT$"}
  }

Here’s the instructions on the webpage: URL: https://blockchain.info/tobtc
Convert x value in the provided currency to btc.
ex: https://blockchain.info/tobtc?currency=USD&value=500
Parameters

  • currency - A currency code. See list above.
  • value - Value to convert.

Returns the value in BTC.

I can use TimeSeriesForecast function but please give me some examples how to use BTC/USD Data to be imported from the above APIs/URLs.

4

This URL:
https://www.quandl.com/data/BITSTAMP/USD-USD-BITCOIN-Exchange-Rate
Is just a webpage. What you want is the URL that uses the quandl API. In this case, you want:

data = Import["https://www.quandl.com/api/v1/datasets/BITSTAMP/USD.csv"]

Please see quandl’s documentation on how to use their API: https://www.quandl.com/help/api

5

Have you checked out the “BTCpredictor” code on github yet? GitHub - panditanvita/BTCpredictor: Bitcoin price prediction algorithm using bayesian regression techniques · GitHub