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
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
| from __future__ import annotations
import matplotlib.pyplot as plt
import numpy as np
from sklearn.datasets import make_regression
from sklearn.linear_model import ElasticNet, ElasticNetCV
from sklearn.metrics import mean_squared_error, r2_score
from sklearn.model_selection import train_test_split
def run_elastic_net_demo(
n_samples: int = 500,
n_features: int = 30,
n_informative: int = 10,
noise: float = 15.0,
duplicate_features: int = 5,
label_scatter_x: str = "predicted",
label_scatter_y: str = "actual",
label_scatter_title: str = "Predicted vs. actual",
label_bar_title: str = "Top coefficients",
label_bar_ylabel: str = "coefficient",
top_n: int = 10,
) -> dict[str, float]:
"""Fit Elastic Net with CV, report metrics, and plot predictions/coefs.
Args:
n_samples: Number of generated samples.
n_features: Total features before duplication.
n_informative: Features with non-zero weights in the generator.
noise: Standard deviation of noise added to targets.
duplicate_features: Number of leading features to duplicate for correlation.
label_scatter_x: Label for the scatter plot x-axis.
label_scatter_y: Label for the scatter plot y-axis.
label_scatter_title: Title for the scatter plot.
label_bar_title: Title for the coefficient bar plot.
label_bar_ylabel: Y-axis label for the coefficient plot.
top_n: Number of largest-magnitude coefficients to display.
Returns:
Dictionary with training/test metrics for inspection.
"""
rng = np.random.default_rng(seed=123)
X, y, coef = make_regression(
n_samples=n_samples,
n_features=n_features,
n_informative=n_informative,
noise=noise,
coef=True,
random_state=123,
)
correlated = X[:, :duplicate_features] + rng.normal(
scale=0.1, size=(X.shape[0], duplicate_features)
)
X = np.hstack([X, correlated])
feature_names = np.array([f"x{i}" for i in range(X.shape[1])], dtype=object)
X_train, X_test, y_train, y_test = train_test_split(
X, y, test_size=0.25, random_state=42
)
enet_cv = ElasticNetCV(
l1_ratio=[0.2, 0.5, 0.7, 0.9, 0.95, 1.0],
alphas=np.logspace(-3, 1, 30),
cv=5,
random_state=42,
max_iter=5000,
)
enet_cv.fit(X_train, y_train)
enet = ElasticNet(
alpha=float(enet_cv.alpha_),
l1_ratio=float(enet_cv.l1_ratio_),
max_iter=5000,
random_state=42,
)
enet.fit(X_train, y_train)
train_pred = enet.predict(X_train)
test_pred = enet.predict(X_test)
metrics = {
"best_alpha": float(enet_cv.alpha_),
"best_l1_ratio": float(enet_cv.l1_ratio_),
"train_r2": float(r2_score(y_train, train_pred)),
"test_r2": float(r2_score(y_test, test_pred)),
"test_rmse": float(mean_squared_error(y_test, test_pred, squared=False)),
}
top_idx = np.argsort(np.abs(enet.coef_))[-top_n:][::-1]
fig, axes = plt.subplots(1, 2, figsize=(12, 5))
ax_scatter, ax_bar = axes
ax_scatter.scatter(test_pred, y_test, alpha=0.6, color="#1f77b4")
ax_scatter.plot(
[y_test.min(), y_test.max()],
[y_test.min(), y_test.max()],
color="#d62728",
linestyle="--",
linewidth=1.5,
)
ax_scatter.set_title(label_scatter_title)
ax_scatter.set_xlabel(label_scatter_x)
ax_scatter.set_ylabel(label_scatter_y)
ax_bar.bar(
np.arange(top_n),
enet.coef_[top_idx],
color="#ff7f0e",
)
ax_bar.set_xticks(np.arange(top_n))
ax_bar.set_xticklabels(feature_names[top_idx], rotation=45, ha="right")
ax_bar.set_title(label_bar_title)
ax_bar.set_ylabel(label_bar_ylabel)
fig.tight_layout()
plt.show()
return metrics
metrics = run_elastic_net_demo(
label_scatter_x="予測値",
label_scatter_y="実測値",
label_scatter_title="予測と実測の比較",
label_bar_title="重要な係数",
label_bar_ylabel="係数の大きさ",
)
print(f"最適な alpha: {metrics['best_alpha']:.4f}")
print(f"最適な l1_ratio: {metrics['best_l1_ratio']:.2f}")
print(f"訓練データのR^2: {metrics['train_r2']:.3f}")
print(f"テストデータのR^2: {metrics['test_r2']:.3f}")
print(f"テストRMSE: {metrics['test_rmse']:.3f}")
|