关于「实证闭环:社群拓扑体检器、人工复核与预注册检验」,《社群拓扑体检》给出的定量结论是什么?

结论:第八章给出社群月度拓扑体检流程、社群拓扑体检器与预注册检验方案(每组 41 个审计社群、复核 93 个名单账号)。书中以标注框区分已发表的证据、可运行的 Python 计算(19 段程序)、原典与作者的假说。

数据口径:书中所有社群与成员都是固定种子的合成数据,不使用任何真实社群或个人的数据;全部计算由书中程序生成,参数为示意;本书是非个人化经营建议,「水军识别」只产出供人工复核的嫌疑名单,不自动处罚任何成员。 第二章用并查集求零维条码、用模 2 列约化求一维条码,在两百个合成社群上,点赞总数的 AUC 只有 0.427,一维条码前三长之和 L3 为 0.975;补量社群的真实核心增至 30 人时降到 0.673,此时须靠零维短条。

English version

对照数据表

下表数字由本章 Python 程序原样打印,可重新运行复现。 程序只用 Python 标准库与结论吻合的打印值: 41, 93

z(0.975) = 1.9600, z(0.80) = 0.8416
AUC H0AUC H1communities per group
0.700.8541
0.700.8098
0.600.8028
0.500.7520
flagged accounts to audit: 93 (half-width 0.0996)
查看完整验算程序
# Chapter 8. p82: Sample sizes for the pre-registered test.
# Primary endpoint: AUC of L3 for telling audited genuine communities
# from audited padded ones. H0: AUC = 0.70 (a generous bar for any
# activity metric); H1: AUC = 0.85. One-sided alpha 0.025, power 0.80,
# equal groups, Hanley and McNeil (1982) variance. Second: how many
# flagged accounts to audit so that a Wilson 95 % interval for the
# precision of the list has half-width <= 0.10 at the worst case 0.5.
import math
from statistics import NormalDist

Z = NormalDist()


def var_auc(a, n):
    q1, q2 = a / (2.0 - a), 2.0 * a * a / (1.0 + a)
    return (a * (1 - a) + (n - 1) * (q1 - a * a)
            + (n - 1) * (q2 - a * a)) / (n * n)


za, zb = Z.inv_cdf(0.975), Z.inv_cdf(0.80)
print("z(0.975) = %.4f, z(0.80) = %.4f" % (za, zb))
print("AUC H0  AUC H1  communities per group")
for a0, a1 in ((0.70, 0.85), (0.70, 0.80), (0.60, 0.80), (0.50, 0.75)):
    n = 5
    while (za * math.sqrt(var_auc(a0, n)) + zb * math.sqrt(var_auc(a1, n))
           > a1 - a0):
        n += 1
    print(" %.2f    %.2f    %d" % (a0, a1, n))
p = 0.5
for m in range(10, 400):
    c = 1 + za * za / m
    half = za * math.sqrt(p * (1 - p) / m + za * za / (4 * m * m)) / c
    if half <= 0.10:
        print("flagged accounts to audit: %d (half-width %.4f)"
              % (m, half))
        break

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《持续同调拓扑贝蒂数、高净值私域社群结构空洞与虚假抱团水军识别母典》第 8 章「实证闭环:社群拓扑体检器、人工复核与预注册检验」(版本 1.0.0). 5D Library. DOI 待发布

DOI 待发布 · 专著页正式发布后上线(本站暂无专著页) · 5DL-B-236 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026236,
  author    = {Fan, Yafu},
  title     = {Persistent Homology and Betti Numbers: Structural Voids in High-Net-Worth Private Communities and the Review of Fake Cliques},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论