Using Data Generator Utilities¶

In this tutorial, we explore some utilities for generating time series data. EERILY provides noise generators (eerily.generators.utils.noises), event generators (eerily.generators.utils.events), a basic stepper to help the users create iterators of time series (eerily.generators.utils.stepper), and finally factories to unify them.

In [1]:

import matplotlib.pyplot as plt

seed = 42
length = 20

import matplotlib.pyplot as plt seed = 42 length = 20

Noise¶

Noise is an important part of a real word dataset. To generate realistic time series data, we always add noise to the data points.

In [2]:

from eerily.generators.utils.noises import GaussianNoise

from eerily.generators.utils.noises import GaussianNoise

GaussianNoise requires a mean and a standard deviation.

In [3]:

gn_mu = 1
gn_std = 0.1
gn = GaussianNoise(mu=gn_mu, std=gn_std, seed=seed)

gn_mu = 1 gn_std = 0.1 gn = GaussianNoise(mu=gn_mu, std=gn_std, seed=seed)

The Gaussian noise gn is an iterator.

In [4]:

gn_data = [next(gn) for _ in range(length)]
gn_data

gn_data = [next(gn) for _ in range(length)] gn_data

Out[4]:

[1.030471707975443,
 0.8960015893759504,
 1.0750451195806456,
 1.0940564716391215,
 0.8048964811346163,
 0.8697820493137682,
 1.0127840403167285,
 0.9683757407656418,
 0.9983198842495711,
 0.914695607242642,
 1.0879397974862828,
 1.0777791935428949,
 1.0066030697561217,
 1.1127241206968033,
 1.0467509342252046,
 0.9140707537116761,
 1.0368750784082499,
 0.9041117399171001,
 1.0878450301307272,
 0.9950074089013747]

In [5]:

fig, ax = plt.subplots(figsize=(10, 6.18))
ax.plot(range(length), gn_data, color="k", marker="o")
ax.set_xlabel("Time Step")
ax.set_ylabel("Noise")
ax.set_title("GaussianNoise Example")

fig, ax = plt.subplots(figsize=(10, 6.18)) ax.plot(range(length), gn_data, color="k", marker="o") ax.set_xlabel("Time Step") ax.set_ylabel("Noise") ax.set_title("GaussianNoise Example")

Out[5]:

Text(0.5, 1.0, 'GaussianNoise Example')

Events¶

Suppose we set a sensor on a road to record whether a car is passing. Assuming the traffic flow is sparse, we expect to see some spikes in this time series. A simple model is a Poisson process.

In this section, we demo the PoissonEvent class.

In [6]:

from eerily.generators.utils.events import PoissonEvent

from eerily.generators.utils.events import PoissonEvent

PoissonEvent requires one argument rate. We can also set the seed for reproducibility.

In [7]:

pe_rate = 0.5

pe_rate = 0.5

In [8]:

pe = PoissonEvent(rate=pe_rate, seed=seed)

pe = PoissonEvent(rate=pe_rate, seed=seed)

The event defined pe is an iterator. We iterate to get some data.

In [9]:

pe_data = [next(pe) for _ in range(length)]
pe_data

pe_data = [next(pe) for _ in range(length)] pe_data

Out[9]:

[0, 1, 0, 0, 1, 0, 0, 0, 1, 1, 1, 0, 0, 0, 1, 1, 0, 1, 0, 0]

In [10]:

fig, ax = plt.subplots(figsize=(10, 6.18))
ax.plot(range(length), pe_data, color="k", marker="o")
ax.set_xlabel("Time Step")
ax.set_ylabel("Event")
ax.set_title("PoissonEvent Example")

fig, ax = plt.subplots(figsize=(10, 6.18)) ax.plot(range(length), pe_data, color="k", marker="o") ax.set_xlabel("Time Step") ax.set_ylabel("Event") ax.set_title("PoissonEvent Example")

Out[10]:

Text(0.5, 1.0, 'PoissonEvent Example')