模型中的机器学习偏差

import numpy as np from scipy.stats import chi2_contingency

def demographic_parity_analysis( total_approval_rate=0.60, female_approval_rate=0.45, male_approval_rate=0.70 ): # 计算人口统计均等性差异 demographic_parity_diff = abs(male_approval_rate - female_approval_rate)

1# 统计显著性检验
2total_sample_size = 1000
3female_sample = np.random.binomial(total_sample_size, female_approval_rate)
4male_sample = np.random.binomial(total_sample_size, male_approval_rate)
5
6# 独立性卡方检验
7contingency_table = np.array([
8    [female_sample, total_sample_size - female_sample],
9    [male_sample, total_sample_size - male_sample]
10])
11chi2, p_value = chi2_contingency(contingency_table)[:2]
12
13return {
14    "demographic_parity_difference": demographic_parity_diff,
15    "chi2_statistic": chi2,
16    "p_value": p_value,
17    "bias_threshold": 0.2,
18    "is_biased": demographic_parity_diff > 0.2
19}

outputs = demographic_parity_analysis() print(outputs)

def equalized_odds_evaluation( group_a_true_positive_rate=0.65, group_a_false_positive_rate=0.35, group_b_true_positive_rate=0.40, group_b_false_positive_rate=0.15 ): # 计算各类率的差异 tpr_disparity = abs(group_a_true_positive_rate - group_b_true_positive_rate) fpr_disparity = abs(group_a_false_positive_rate - group_b_false_positive_rate)

1return {
2    "true_positive_rate_disparity": tpr_disparity,
3    "false_positive_rate_disparity": fpr_disparity,
4    "total_disparity": tpr_disparity + fpr_disparity,
5    "is_equalized": tpr_disparity < 0.1 and fpr_disparity < 0.1
6}

outputs = equalized_odds_evaluation() print(outputs)