huncholane/time-series-cookbookPublic

main

time-series-cookbook/ch1/1.06_computing_autocorrelation.py

41 lines1.1 KBPython

RawDownloadHistory

1	"""
2	# Computing autocorrelation
3	Correlation against itself at different lags to quantify how past values
4	effect the future.
5
6	Partial correlation has more control by comparing at shorter lags
7	"""
8	import os
9	from statsmodels.graphics.tsaplots import plot_acf, plot_pacf
10	import pandas as pd
11	import matplotlib.pyplot as plt
12
13	# Chart config
14	plt.rcParams['figure.figsize'] = (12, 6)
15
16	# Ensure assets directory
17	os.makedirs("assets", exist_ok=True)
18
19	# Load solar data
20	data = pd.read_csv(
21	"../assets/datasets/time_series_solar.csv",
22	parse_dates=["Datetime"],
23	index_col="Datetime",
24	)
25	series = data["Incoming Solar"]
26
27	# Resample to daily
28	series_daily = series.resample("D").sum()
29
30	# Compute autocorrelation, acf = Auto Correlation Function
31	# Compares against lags up to a year away
32	# acf_scores = acf(x=series_daily, nlags=365)
33
34	# Compute partial autocorrelation
35	# pacf_scores = pacf(x=series_daily, nlags=365)
36
37	# Convience functions to do math and plot
38	plot_acf(series_daily, lags=365)
39	plt.savefig("assets/autocorrelation.png")
40	plot_pacf(series_daily, lags=365)
41	plt.savefig("assets/partion_autocorrelation.png")