relief特征选择

import numpy as np
from sklearn.datasets import make_classification
from sklearn.preprocessing import normalize

def Relief(features, labels, iter_ratio=100):
    m, n = np.shape(features)
    weight = np.zeros(n)
    for _ in range(iter_ratio):
        for i in range(m):
            index_i = np.random.randint(0, m)
            sample_R = features[index_i]
            near_hits = []
            near_misses = []
            for j in range(n):
                if features[index_i, j] != 0:
                    continue
                diff = np.abs(sample_R - features[index_i])
                if labels[index_i] == labels[np.argmin(diff)]:
                    near_hits.append(np.argmin(diff))
                else:
                    near_misses.append(np.argmin(diff))
            for j in range(n):
                if j not in near_hits:
                    weight[j] += np.sum(diff[near_misses == j])
                elif j not in near_misses:
                    weight[j] -= np.sum(diff[near_hits == j])
    weight /= len(near_hits)
    return weight

# 示例数据集
X, y = make_classification(n_samples=1000, n_features=20, n_informative=2, n_redundant=10, random_state=42)
# 特征缩放
X = normalize(X)
# 计算特征权重
feature_weights = Relief(X, y)
print(feature_weights)