auc计算逻辑

最新推荐文章于 2025-07-18 10:48:39 发布

原创

最新推荐文章于 2025-07-18 10:48:39 发布 · 439 阅读

0 ·

CC 4.0 BY-SA版权

文章标签：

#auc #udf

该博客展示了如何在Python和Java中实现AUC（Area Under the Curve）的计算，用于评估分类模型的性能。提供了详细的代码示例，包括使用sklearn库的roc_auc_score函数以及自定义的AUC计算函数，适用于大数据量的情况。

auc计算逻辑

实现auc，python

# -*- coding utf-8 -*-
# @Author: 
# @Date: 2021/3/3 11:07 上午


from sklearn.metrics import roc_auc_score
import time


def auc(y_true, y_score):
    assert len(y_true) == len(y_score)

    m = sum(y_true)
    n = len(y_true) - m

    bins = 200
    bin_width = 1 / bins
    p_bins = [0 for _ in range(bins)]
    n_bins = [0 for _ in range(bins)]

    # 分桶
    for i in range(m + n):
        index = int(y_score[i] / bin_width)
        if y_true[i] == 1:
            p_bins[index] += 1
        else:
            n_bins[index] += 1

    accumulated_n = 0
    pair = 0

    for i in range(bins):
        pair += p_bins[i] * accumulated_n + p_bins[i] * n_bins[i] * 0.5
        # 滞后一位
        accumulated_n += n_bins[i]

    return pair / (m * n)


label = [1, 0, 1, 1, 0, 1, 0, 0, 1]
score = [0.5, 0.