tobiasraabe / crypto Goto Github PK

Every trade on Poloniex trade history has a unique tradeID. 8631 entries are missing. Investigate using src/exploration/poloniex.ipynb to track down the missing values. It seems that the missing values are a missing period and not randomly distributed.

• Are the values available with the API?
• Are they lost during the validation phase?

Furthermore, the validation of periods is poorly designed and more an adhoc method. If there is an easy improvement, implement it!

Sag bloß du probierst jetzt org-mode aus. ;-)

Sag mir mal zuerst, was ich machen soll. Ich bin übrigens 02.01. wieder in Bonn.

This issue is a reference list for interesting information which may help to make that project better.

Analysis:

• https://gmarti.gitlab.io/cryptocurrency/2017/08/25/download-cryptocoins-api-python.html
• https://blog.patricktriest.com/analyzing-cryptocurrencies-python/
• http://pythondata.com/forecasting-time-series-data-with-prophet-part-1/

Data:

• https://github.com/kennethreitz/awesome-coins
• https://github.com/kennethreitz/coinbin.org
• https://github.com/dashee87/cryptory

Docs:

• https://raw.githubusercontent.com/jupyterlab/jupyterlab/master/docs/source/index.rst

Description

What I have not recognized so far is that pickle objects are just a string representation of the python object and get extremely big extremely fast.

For example, the combined datasets of chart data and trade history for the BTC_POT exchange with 20 lags is about 1gb big. Compressing the file with 7z yields a 10mb file. An export to .csv decrease the file size only by a factor of ten.

The problem exacerbates if we pick a more realistic approach and use 100 lags.