Financial toolbox

The financial toolbox regroups a collection of useful functions to process cryptocurrencies and other market data. It lets you import market data from different sources (currently only CSV), create meaningful features from a market dataset to feed machine learning algorithmes, and visualize those data.

This toolbox is currently in development and more functions will be added time to time.

Installation

Simply download the pytorency folder and copy into in your project.

Import market data

Market candles can be imported from CSV files. Times are supported either as timestamps or as text.

import fintoolbox as ft
import fintoolbox.load

market = ft.load.market_from_csv('samples/BTC-ETH.csv', sep=';')
market.head()

The above code produces the following pandas dataframe where time is a DatetimeIndex.

                         open      high       low     close  base_volume       volume
time
2017-10-31 00:00:00  0.050000  0.050145  0.049845  0.050000    52.650524  1052.762789
2017-10-31 01:00:00  0.050000  0.050145  0.050000  0.050000    54.579999  1090.576553
2017-10-31 02:00:00  0.050000  0.050435  0.050000  0.050208    94.177311  1876.060180
2017-10-31 03:00:00  0.050208  0.050389  0.050016  0.050210    65.767919  1308.115839
2017-10-31 04:00:00  0.050200  0.050400  0.050010  0.050200    64.754838  1288.355149

Transform market data

Market resampling

Let say you have an hourly dataset but you won't need that much precision. The function downscale_market can resample a market dataframe. In the example below, we downscale hourly candles to 2-day candles.

import fintoolbox as ft
import fintoolbox.transform

market_2d = ft.transform.downscale_market(market, freq='2d')
market_2d.head()

                open      high       low    close   base_volume         volume
time                                                                          
2017-10-31  0.050000  0.050435  0.042800  0.04286   5824.147753  124559.619790
2017-11-02  0.042940  0.043475  0.037800  0.04260   9107.473992  222773.889780
2017-11-04  0.042600  0.043120  0.039000  0.04000   3727.084295   92015.885006
2017-11-06  0.039918  0.043407  0.039820  0.04100   5202.159551  125317.977368
2017-11-08  0.041000  0.046174  0.038119  0.04486  11638.793260  272094.869171

Now each row of the dataframe represents 2 days instead of 1 hour.

Add variations

Variation columns can be easily calculated with the add_variation function. A variation is the ration of prices at two different epochs. The following code adds the variation between the current on the upcoming epoch (from N+0 to N+1), as well as the variation over the 3 last epochs (from N-3 to N).

By default, variations are calculated on close prices. This can be changed with the column parameter.

ft.transform.add_variation(market_2d, 0, 1)
ft.transform.add_variation(market_2d, -3, 0)
market_2d.head()

                open      high       low    close   ...   var_N_N+1  var_N-3_N
time                                                                                  
2017-10-31  0.050000  0.050435  0.042800  0.04286   ...    0.993934        NaN
2017-11-02  0.042940  0.043475  0.037800  0.04260   ...    0.938967        NaN
2017-11-04  0.042600  0.043120  0.039000  0.04000   ...    1.025000        NaN
2017-11-06  0.039918  0.043407  0.039820  0.04100   ...    1.094144   0.956603
2017-11-08  0.041000  0.046174  0.038119  0.04486   ...    1.105325   1.053050

Plotting candlesticks

Prices candlesticks and volume bars can be plotted using Matplotlib with a single line of code :

import fintoolbox.plotting
ft.plotting.plot_candles(market_2d)

aurelien-roy / financial-toolbox Goto Github PK