关于「冷启动:用均值与品类先验接住新顾客与新商品」,《稀疏购买记录的偏好补全》给出的定量结论是什么?

结论:第六章讲冷启动,新顾客有八条记录时向品类先验收缩的折入法误差为 0.2434。

数据口径:书中顾客数、商品数、口味维度与购买概率均为示意参数;全部计算由书中程序生成;本书属于非个人化经营建议,推荐流程只用店铺自有数据,不外传个人数据,不做跨平台追踪。 第七章考辨《周易·系辞上》「方以类聚,物以群分」与斯皮尔曼 1904 年的一般因素,指出纯分群与混合成员下两种模型各有胜负。

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: 0.2434

20 new customers; RMSE on their unobserved products
kglobalproductsegmentfold-in
00.84020.72830.48350.4835
10.83740.72940.48320.4313
20.85220.74230.48680.4207
40.85400.74440.48870.3586
80.86630.74400.48180.2434
查看完整验算程序
# Chapter 6. p61: a new customer with k purchases.
# Model: 60 established customers x 24 products in 4 shelf categories;
# customers fall in 4 segments by first-purchase category, and their
# rank-3 tastes scatter around a segment mean (illustrative). Product
# factors come from ALS on the established customers (40 percent
# observed). For 20 new customers we predict every product from k
# observed scores: global mean; product means; segment prior; and a
# ridge fold-in shrunk toward the segment prior (lambda 2).
import math
import random


def dot(a, b):
    return math.fsum(x * y for x, y in zip(a, b))


def solve(A, b):
    # Gaussian elimination with partial pivoting for a small system
    n = len(b)
    M = [list(A[i]) + [b[i]] for i in range(n)]
    for k in range(n):
        p = max(range(k, n), key=lambda i: abs(M[i][k]))
        M[k], M[p] = M[p], M[k]
        for i in range(k + 1, n):
            f = M[i][k] / M[k][k]
            M[i] = [M[i][j] - f * M[k][j] for j in range(n + 1)]
    x = [0.0] * n
    for i in range(n - 1, -1, -1):
        s = math.fsum(M[i][j] * x[j] for j in range(i + 1, n))
        x[i] = (M[i][n] - s) / M[i][i]
    return x


def update(trip, F, r, lam, prior=None):
    # weighted ridge step for one row: minimise
    # sum w (y - f.x)^2 + lam |x - prior|^2 over x
    pr = prior if prior is not None else [0.0] * r
    A = [[math.fsum(w * F[j][a] * F[j][b] for j, _, w in trip)
          + (lam if a == b else 0.0) for b in range(r)]
         for a in range(r)]
    y = [math.fsum(w * F[j][a] * v for j, v, w in trip) + lam * pr[a]
         for a in range(r)]
    return solve(A, y)


def als(nu, ni, obs, r, lam, iters, seed, V=None, trace=None):
    # alternating least squares on (u, i, value, weight) tuples
    rng = random.Random(seed)
    if V is None:
        V = [[rng.gauss(0.0, 1.0) for _ in range(r)] for _ in range(ni)]
    bu = [[] for _ in range(nu)]
    bi = [[] for _ in range(ni)]
    for u, i, x, w in obs:
        bu[u].append((i, x, w))
        bi[i].append((u, x, w))
    U = None
    for t in range(iters):
        U = [update(bu[u], V, r, lam) for u in range(nu)]
        V = [update(bi[i], U, r, lam) for i in range(ni)]
        if trace is not None:
            trace.append((t + 1, U, V))
    return U, V


def factors(n, r, rng):
    return [[rng.gauss(0.0, 1.0) for _ in range(r)] for _ in range(n)]


def product(A, B):
    return [[dot(a, b) for b in B] for a in A]


def mask(nu, ni, p, rng):
    return [(u, i) for u in range(nu) for i in range(ni)
            if rng.random() < p]

def store(rng, nu=60, ni=24, r=3, segs=4):
    # customer segment = first purchase category (store data);
    # product category = shelf; latent factors share a segment mean
    cen = [[rng.gauss(0, 1) for _ in range(r)] for _ in range(segs)]
    cat = [[rng.gauss(0, 1) for _ in range(r)] for _ in range(segs)]
    sg = [u % segs for u in range(nu)]
    ct = [i % segs for i in range(ni)]
    A = [[c + rng.gauss(0, 0.5) for c in cen[sg[u]]] for u in range(nu)]
    B = [[c + rng.gauss(0, 0.5) for c in cat[ct[i]]] for i in range(ni)]
    T = [[3.0 + 0.4 * dot(a, b) for b in B] for a in A]
    return T, sg, ct


rng = random.Random(20261113)
T, sg, ct = store(rng, nu=80)
nu, ni, r = 60, 24, 3
seen = mask(nu, ni, 0.4, rng)
g = math.fsum(T[u][i] for u, i in seen) / len(seen)
obs = [(u, i, T[u][i] - g, 1.0) for u, i in seen]
U, V = als(nu, ni, obs, r, 1.0, 60, 6)
im = []
for i in range(ni):
    xs = [T[u][j] for u, j in seen if j == i]
    im.append(math.fsum(xs) / len(xs))
pri = []
for s in range(4):
    us = [U[u] for u in range(nu) if sg[u] == s]
    pri.append([math.fsum(x[k] for x in us) / len(us) for k in range(r)])
new = list(range(60, 80))
print("20 new customers; RMSE on their unobserved products")
print("   k   global   product   segment   fold-in")
for k in (0, 1, 2, 4, 8):
    e = {"g": [], "p": [], "s": [], "f": []}
    for u in new:
        order = list(range(ni))
        random.Random(1000 + u).shuffle(order)
        known, rest = order[:k], order[k:]
        pr = pri[sg[u]]
        trip = [(i, T[u][i] - g, 1.0) for i in known]
        uf = update(trip, V, r, 2.0, prior=pr)
        for i in rest:
            e["g"].append((g - T[u][i]) ** 2)
            e["p"].append((im[i] - T[u][i]) ** 2)
            e["s"].append((g + dot(pr, V[i]) - T[u][i]) ** 2)
            e["f"].append((g + dot(uf, V[i]) - T[u][i]) ** 2)
    out = [math.sqrt(math.fsum(e[x]) / len(e[x])) for x in "gpsf"]
    print("  {:2d}   {:.4f}   {:.4f}    {:.4f}    {:.4f}".format(k, *out))

唯一数据源与引用

范雅福(Fan, Y.). (2026). 《低秩矩阵补全与交替最小二乘、稀疏购买记录推断未表态偏好母典》第 6 章「冷启动:用均值与品类先验接住新顾客与新商品」(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-328 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026328,
  author    = {Fan, Yafu},
  title     = {Low-Rank Matrix Completion and Alternating Least Squares: Inferring Unstated Preferences from Sparse Purchase Records},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

本书其他章节结论