Datathons SolutionsLearnTeam solutions

Prediction Model for Crypto Currency in R

Bitcoins are cryptocurrency systems, that enable its users to exchange payments without passing through a central authority (Eg. Reserve Bank of India, Federal Bank etc). They were developed in 2008, using the Blockchain Technology. In the present article, methods to create prediction models have been implemented. The model considers a sample data of 3 months spaced over 5 minutes for each day. The Training data and Testing data are developed on that dataset for twenty bitcoins; viz: Bitcoin, Bitcoin Cash, Bitcoin Gold, Cardeno, Dash, Dogecoin, Eos, Ethereum, Ethereum Class, Iota, Lisk, Litrcoin, Monero, NEMcoin, Neo, Ripple, Stellar, Tether, Tron, Zcash.
The prediction models used are ARIMA, Exponential Smoothing and Neural Networks on R. The models calculate the values for the next time instant, i.e. next five minutes and the code developed goes on continuing it (predicting next 5-minute price) for all 288 time-points in a day.


2 thoughts on “Prediction Model for Crypto Currency in R

  1. 0

    1. You say there are missing timestamps but are there missing values?
    2. How did you imputed the missing values?
    3. Did you find a trend/seasonality in the data provided?
    4. How good is your predictive model? What are the metrics you have used to evaluate it?
    5. Would you bet your own money on your predictions? If so how much?

  2. 0

    Your code is self sufficient (i.e. all inputs are present or part of the task supplied) which is a big plus. That said, there seems to be issues with the code, which makes it executable ( for example, variable pred_et is never defined)

    In general, the provided code would not give predictions for the required dates:
    30.01.2018, 06.02.2018, 20.02.2018, 09.03.2018, 18.03.2018

    However, it seems that it can be changed to do that.

    You are missing one of the currencies, because of what is most probably typo – the last selected currency is “146” rather than the required “1465”.

    I like that you have formed three different predictions, but you do not try to explain which one is better – even simple comparisons between the three would increase the usefulness of the output.
    Linked with this, the statistical tests are missing that check goodness of the predictions.

Leave a Reply