1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
| import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
rng = np.random.default_rng(19)
dates = pd.date_range("2020-01-01", periods=240, freq="D")
segments = [
(0, 80, 1.5),
(70, 95, 1.8),
(140, 110, 2.2),
]
values = np.empty(len(dates))
segment_bounds = [seg[0] for seg in segments[1:]] + [len(dates)]
for (start, mean, scale), end in zip(segments, segment_bounds):
values[start:end] = mean + rng.normal(0, scale, end - start)
series = pd.Series(values, index=dates)
rolling = series.rolling(window=14, min_periods=1).mean()
fig, ax = plt.subplots(figsize=(7, 4))
ax.plot(series.index, series, color="#64748b", alpha=0.7, label="観測値")
ax.plot(rolling.index, rolling, color="#2563eb", linewidth=2, label="14日移動平均")
for start, _, _ in segments[1:]:
ax.axvline(dates[start], color="#ef4444", linestyle="--", linewidth=1.5)
ax.set_title("平均の変化点を確認")
ax.set_xlabel("日付")
ax.set_ylabel("値")
ax.legend()
ax.grid(alpha=0.3)
fig.tight_layout()
fig.savefig("static/images/timeseries/change_point.svg")
|
