• Series2GAF can be used to transform time series into Gramian Angular Field.
• There are many useful utility features for financial research in Series2GAF, and will be mentioned in the following sections.

simple time series encoding package, focused on financial tasks.

this is an simple example:

import numpy as np
from series2gaf import GenerateGAF

# create a random sequence with 200 numbers
# all numbers are in the range of 50.0 to 150.0
random_series = np.random.uniform(low=50.0, high=150.0, size=(200,))

# set parameters
timeSeries = list(random_series)
windowSize = 50
rollingLength = 10
fileName = 'demo_%02d_%02d'%(windowSize, rollingLength)

# generate GAF pickle file (output by Numpy.dump)
GenerateGAF(all_ts = timeSeries,
            window_size = windowSize,
            rolling_length = rollingLength,
            fname = fileName)

now we get a file named demo_50_10_gaf.pkl in current directory. inside the pickle file, you got a grammian angular field with shape (15, 50, 50).

• shape[0] refers to data amount : floor((len(timeSeries)-(normalize_window_scaling-1)*windowSize)/windowSize)
• shape[1] refers to image width : windowSize
• shape[2] refers to image height : windowSize

Thats It!

if we want to preview the image results, just add these to the code:

from series2gaf import PlotHeatmap

gaf = np.load('%s_gaf.pkl'%fileName)
PlotHeatmap(gaf)

we can now find GAF heapmap images in a new child directory /output_img!

def GenerateGAF(all_ts, window_size, rolling_length, fname,
                normalize_window_scaling=1.0, method='summation', scale='[0,1]'):
  ...
  
def PlotHeatmap(all_img, save_dir='output_img'):
  ...

• all_ts: list
      the time series we want to transform.

• window_size: int
      the sliding window size for transforming sequences into GAF images

• rolling_length: int
      also known as "stride value" for the sliding window

• fname: str
      output file name, the output pickle file will be named as "[fname]_gaf.pkl"

• normalize_window_scaling: float, optional
      default: 1.0
      normalize the values in the windows, but considering a ratio of previous values

• method: str, optional
      default: 'summation'
      'summation' is for GASF ( calculate cos(x1+x2) )
      'difference' if for GADF ( calculate sin(x1-x2) )

• scale: str, optoinal
      default: '[0,1]'
      '[0,1]' means normalize the sequence in the range of 0 and 1
      '[-1,1]' means normalize the sequence in the range of -1 and 1

• all_img: numpy.array
      input GAF multi-dimension array

• save_dir: str, optional
      default: 'output_img'
      directory for output images