Hands-On Machine Learning for Cybersecurity

Partial autocorrelation function

Partial autocorrelation function (PACF) can be defined as a time series where there is a restricted or incomplete correlation between the values for shorter time lags.

PACF is not at all like ACF; with PACE the autocorrelation of a data point at the current point and the autocorrelation at a period lag have a direct or indirect correlation. PACF concepts are heavily used in autoregressive models.

In Python, the PACF function can be computed as follows:

import matplotlib.pyplot as pltimport numpy as npimport pandas as pfrom statsmodels.graphics.tsaplots import plot_pacfdata = p.Series(0.7 * np.random.rand(1000) + 0.3 * np.sin(np.linspace(-9 * np.pi, 9 * np.pi, num=1000)))plot_pacf(data, lag = 50)pyplot.show() ...