Tabular Time Series

Summary

This repo was created as I did not find a function able to transform a time-series (1D) into a tabular format (X, y).

Usage

TimeSeriesGenerator

The docstring is as follows. Given a 1D array data = [0, 1, 2, 3, 4, 5, 6], generates X, y following the parameters p(autoregressive), s (seasonal) and n (lenght of y).

Therefore, it makes it possible to train a neural network (e.g.) that 2 autoregressive entries (e.g. p = 2) and predicts the next two (n = 2) using 2 (n = 2) entries with lag 4 (s = 4).

>> data = [0, 1, 2, 3, 4, 5, 6]
>> p, n = 2, 2
>> ts = TimeSeriesGenerator(data, p, n)
>> for _, X, y in ts:
...    print(X, y)
    [0, 1] [2, 3]
    [1, 2] [3, 4]
    [2, 3] [4, 5]
    [3, 4] [5, 6]
>> p, n, s = 2, 2, 4
>> ts = TimeSeriesGenerator(data, p, n, s)
>> for X, y in ts:
...    # both y have their respective seasonal entry
...    print(data.index(y[0]) - data.index(X[0]) == s, data.index(y[1]) - data.index(X[1]) == s)
...    print(X, y)
    [0, 1], [2, 3] [4, 5]
    [1, 2], [3, 4] [5, 6]

TimeSeriesGeneratorOnline

To support online learning (and streaming) applications, TimeSeriesGeneratorOnline enables applications to give real time measurements and returns a bool b stating if it was possible to generate features, considering the given seasonal s, autoregressive ar and output y.

>>> from tabular_time_series.tsgeneratoronline import TimeSeriesGeneratorOnline
>>> data = [i for i in range(10)]
>>> p, n, s = 2, 2, 4
>>> tsgo = TimeSeriesGeneratorOnline(p, n, s)
>>> for X in data:
...     b, (s, ar, y) = tsgo(X)
...     print(X, '|', b, s, ar, y)
...
0 | False None None None
1 | False None None None
2 | False None None None
3 | False None None None
4 | False None None None
5 | False None None None
6 | True [0, 1] [2, 3] [4, 5]
7 | True [1, 2] [3, 4] [5, 6]
8 | True [2, 3] [4, 5] [6, 7]
9 | True [3, 4] [5, 6] [7, 8]

timeseries2df

Considering that many times a batch array is needed for training, timeseries2df can be used to generate a pandas DataFrame that will contain columns in the format:

>>> from tabular_time_series.tsdf import timeseries2df
>>> data = list(range(10))
>>> p, n, s = 2, 2, 4
>>> df = timeseries2df(data, p, n, s)
>>> df
   y(ts4)_1  y(ts4)_2  y(t-1)  y(t-0)  y(t+1)  y(t+2)
0         0         1       2       3       4       5
1         1         2       3       4       5       6
2         2         3       4       5       6       7
3         3         4       5       6       7       8
4         4         5       6       7       8       9