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
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
| """
熵基多类逻辑回归分类器
检测清醒/低BAC/高BAC三分类
"""
import numpy as np
from sklearn.linear_model import LogisticRegression
from sklearn.model_selection import LeaveOneOut, cross_val_score
from sklearn.metrics import roc_auc_score, classification_report
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline
class AlcoholImpairmentClassifier:
"""
酒精损伤运动学分类器
特征：7个关键运动学特征的 PE 和 SD
分类器：多类逻辑回归
"""
def __init__(self):
self.key_features = [
'ax_pe', 'ax_sd',
'ay_pe', 'ay_sd',
'gx_pe', 'gx_sd',
'throttle_pe',
]
self.pipeline = Pipeline([
('scaler', StandardScaler()),
('clf', LogisticRegression(
multi_class='multinomial',
solver='lbfgs',
C=1.0,
max_iter=1000
))
])
def extract_features(self, imu_data: np.ndarray,
throttle_data: np.ndarray,
window_size: int = 100) -> np.ndarray:
"""
提取特征向量
Args:
imu_data: (N, 6) IMU数据
throttle_data: (N,) 油门位置
window_size: 窗口大小
"""
features = compute_kinematic_features(imu_data, window_size)
num_windows = len(throttle_data) // window_size
pe_values = []
sd_values = []
for w in range(num_windows):
window = throttle_data[w * window_size:(w + 1) * window_size]
pe_values.append(permutation_entropy(window, order=3))
sd_values.append(np.std(window))
features['throttle_pe'] = np.mean(pe_values)
features['throttle_sd'] = np.mean(sd_values)
feat_vector = np.array([
features.get(f, 0.0) for f in self.key_features
])
return feat_vector
def train(self, X: np.ndarray, y: np.ndarray) -> dict:
"""
训练分类器（留一被试交叉验证）
Args:
X: (n_samples, n_features)
y: (n_samples,) 0=清醒, 1=低BAC, 2=高BAC
Returns:
metrics: 准确率和AuROC
"""
loo = LeaveOneOut()
y_pred = np.zeros(len(y))
y_proba = np.zeros((len(y), 3))
for train_idx, test_idx in loo.split(X):
X_train, X_test = X[train_idx], X[test_idx]
y_train, y_test = y[train_idx], y[test_idx]
from sklearn.base import clone
clf = clone(self.pipeline)
clf.fit(X_train, y_train)
y_pred[test_idx] = clf.predict(X_test)
y_proba[test_idx] = clf.predict_proba(X_test)
accuracy = np.mean(y_pred == y)
from sklearn.preprocessing import label_binarize
y_bin = label_binarize(y, classes=[0, 1, 2])
auc = roc_auc_score(y_bin, y_proba, multi_class='ovr', average='weighted')
mask = (y == 0) | (y == 2)
if np.sum(mask) > 1:
auc_sober_high = roc_auc_score(
y[mask] // 2,
y_proba[mask, 2]
)
else:
auc_sober_high = 0.0
self.pipeline.fit(X, y)
return {
'accuracy': accuracy,
'auc_weighted': auc,
'auc_sober_vs_high': auc_sober_high,
'classification_report': classification_report(y, y_pred)
}
if __name__ == "__main__":
classifier = AlcoholImpairmentClassifier()
np.random.seed(42)
n_samples = 75
n_features = len(classifier.key_features)
sober_X = np.random.randn(25, n_features) * 0.5 + np.array([0.8, 0.3] * 3 + [0.7])
low_X = np.random.randn(25, n_features) * 0.5 + np.array([0.65, 0.5] * 3 + [0.6])
high_X = np.random.randn(25, n_features) * 0.5 + np.array([0.45, 0.8] * 3 + [0.45])
X = np.vstack([sober_X, low_X, high_X])
y = np.array([0]*25 + [1]*25 + [2]*25)
results = classifier.train(X, y)
print("=== 分类结果 ===")
print(f"准确率: {results['accuracy']:.1%}")
print(f"加权AuROC: {results['auc_weighted']:.2f}")
print(f"清醒vs高BAC AuROC: {results['auc_sober_vs_high']:.2f}")
print(f"\n{results['classification_report']}")
|