关于「爆款是余震:分支比与 1/(1-n) 乘数」,《评论裂变分支比》给出的定量结论是什么?
结论:把评论看成族谱以后,「热度」被拆成外来速率 \(\mu\) 与分支比 \(n\) 两个可以分别测量的数。每一条外来评论最终带来的评论总数期望是 \(1/(1-n)\);\(n\) 从 0.8 到 0.9,乘数从 5.0 翻到 10.0。同样的 \(n\) 之下,帖子规模的差异可以大到好几倍,最大的一成帖子装下全部评论的四分之一以上,这些差异与内容质量无关。已发表的视频观看研究支持自激级联的框架,并发现大多数内容根本不爆,爆发的衰减形态随分支比变化。下一章转向时间:同样的余震,是被外部一次推起来的,还是自己一点点长起来的,衰减速率与峰形会说出来。
数据口径:书中的帖子、账号、流量、成本与评论价值均为示意参数;全部计算由书中程序生成;本书是非个人化经营建议,只讨论真实用户的评论与平台官方工具买来的曝光,不涉及任何制造评论或虚构稀缺的手段。 第四章用小崎递推、补偿子恒等式与黄金分割求最大似然估计,示意帖的分支比 95 % 区间从 0.42 到 0.92,两小时估计的标准差约 0.186。
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: 0.8, 0.9, 10.0, 5.0
| n | E[K] | sd(K) | P(K=1) | P(K>=20) | share in K>=20 |
| 0.30 | 1.43 | 0.94 | 0.7408 | 0.00000 | 0.0000 |
| 0.50 | 2.00 | 2.00 | 0.6065 | 0.00082 | 0.0097 |
| 0.70 | 3.33 | 5.09 | 0.4966 | 0.01947 | 0.1740 |
| 0.80 | 5.00 | 10.00 | 0.4493 | 0.05167 | 0.3966 |
| 0.90 | 10.00 | 30.00 | 0.4066 | 0.10610 | 0.6939 |
| 0.95 | 20.00 | 87.18 | 0.3867 | 0.14117 | 0.8504 |
| at n = 0.8: 45% of seeds stay alone, yet the mean is 5.0 | |||||
| multiplier 1/(1-n) doubles from n=0.8 to n=0.9: 5.0 -> 10.0 |
查看完整验算程序
# Chapter 1. p11: Cascade size from one outside visitor, by branching ratio.
# Model: every comment triggers a Poisson(n) number of direct replies or
# copycat comments (the Hawkes cluster representation). The total size K
# of one cascade, the seed included, then follows the Borel law
# P(K = k) = exp(-n k) (n k)^(k - 1) / k!, E[K] = 1 / (1 - n),
# Var[K] = n / (1 - n)^3.
# Exact sums, no simulation. Branching ratios are illustrative.
import math
def borel_pmf(k, n):
return math.exp(-n * k + (k - 1) * math.log(n * k)
- math.lgamma(k + 1))
def table_row(n, kmax=20000):
pk = [borel_pmf(k, n) for k in range(1, kmax + 1)]
mean = 1.0 / (1.0 - n)
sd = math.sqrt(n / (1.0 - n) ** 3)
p1 = pk[0]
tail20 = 1.0 - math.fsum(pk[:19])
# share of all comments that sit in cascades of size 20 or more
big = math.fsum((k + 1) * pk[k] for k in range(19, kmax)) / mean
return mean, sd, p1, tail20, big
print("n E[K] sd(K) P(K=1) P(K>=20) share in K>=20")
for n in (0.3, 0.5, 0.7, 0.8, 0.9, 0.95):
mean, sd, p1, tail20, big = table_row(n)
print("%.2f %6.2f %6.2f %.4f %.5f %.4f"
% (n, mean, sd, p1, tail20, big))
n = 0.8
print("at n = %.1f: %.0f%% of seeds stay alone, yet the mean is %.1f"
% (n, 100 * math.exp(-n), 1 / (1 - n)))
print("multiplier 1/(1-n) doubles from n=0.8 to n=0.9: %.1f -> %.1f"
% (1 / (1 - 0.8), 1 / (1 - 0.9)))唯一数据源与引用
范雅福(Fan, Y.). (2026). 《霍克斯自激点过程、评论裂变分支比临界值与直播跟风抢购余震半衰期母典》第 1 章「爆款是余震:分支比与 1/(1-n) 乘数」(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-234 · v1.0.0 · CC-BY-NC-ND-4.0
BibTeX
@book{fan2026234,
author = {Fan, Yafu},
title = {Hawkes Self-Exciting Point Processes: Branching-Ratio Criticality in Comment Cascades and the Aftershock Half-Life of Livestream Herd Buying},
year = {2026},
publisher = {5D Library},
edition = {1.0.0},
note = {DOI pending}
}