ch1/bitcoin_seasonality.py at main · huncholane/time-series-cookbook

time-series-cookbook/ch1/bitcoin_seasonality.py

52 lines1.5 KBPython

1	"""
2	Personal interest to see if bitcoin has seasonality.
3	"""
4
5	import os
6	from statsmodels.graphics.tsaplots import plot_acf
7	import pandas as pd
8	import matplotlib.pyplot as plt
9	from statsmodels.tsa.seasonal import STL
10
11	# Chart config
12	plt.rcParams["figure.figsize"] = (12, 6)
13
14	# Ensure assets directory
15	os.makedirs("assets", exist_ok=True)
16
17	# Load solar data
18	data = pd.read_csv(
19	"../assets/btc_yearly.csv",
20	parse_dates=["Date"],
21	index_col="Date",
22	thousands=",",
23	converters={
24	"Change %": lambda x: float(x.rstrip("%")) / 100,
25	"Vol.": lambda x: pd.to_numeric(x.replace("K", "000").replace("M", "000000").replace('B', '000000000') if x != '' else pd.NA),
26	},
27	)
28	# cutoff = data.index.max() - pd.DateOffset(years=5) # pyright: ignore[reportOperatorIssue]
29	# series = data["Price"]
30	series = data["Price"].iloc[:365*8]
31	# print(data['vol'])
32
33	# Resample to daily
34	series_daily = series.resample("D").sum()
35
36	# Compute autocorrelation, acf = Auto Correlation Function
37	# Compares against lags up to a year away
38	# acf_scores = acf(x=series_daily, nlags=365)
39
40	# Compute partial autocorrelation
41	# pacf_scores = pacf(x=series_daily, nlags=365)
42
43	# Convience functions to do math and plot
44	plot_acf(series_daily, lags=365)
45	plt.savefig("assets/btc_autocorrelation.png")
46
47	result = STL(endog=series_daily, period=365).fit()
48	result.plot()
49	plt.savefig("assets/btc_stl_decomposition.png")
50
51	# plot_pacf(series_daily, lags=365)
52	# plt.savefig("assets/partion_autocorrelation.png")