如何在Pandas中逐行从列表列抽取不重复的随机志愿者值

实现思路 #

需求本质是求学校和志愿者之间的随机完美二分匹配，要求每个学校抽到的志愿者属于自身可选列表，且所有志愿者不重复。我们优先处理可选志愿者数量少的学校，避免后续分配无可用名额，每次抽取后标记志愿者已占用，即可实现需求。

无额外依赖实现方案 #

import random
import pandas as pd

# 原始DataFrame
df = pd.DataFrame({
    'poc': ["a", "b", "c", "d"],
    'school': ["school1", "school2", "school3", "school4"],
    'volunteers': [["sam", "mat", "ali", "mike", "guy", "john"],
                   ["sam", "mat", "ali", "mike"],
                   ["rose", "sam", "mike", "jorge"],
                   ["susan", "jack", "alex", "mat", "mike"]]
})

# 按可选志愿者数量从小到大排序，优先处理可选少的学校减少分配冲突
candidate_list = sorted(df['volunteers'].items(), key=lambda x: len(x[1]))
used_volunteers = set()
match_map = {}

for idx, vol_list in candidate_list:
    # 过滤已被占用的志愿者
    available = [v for v in vol_list if v not in used_volunteers]
    if not available:
        raise ValueError("当前数据不存在可行的不重复分配方案，请检查志愿者可选列表")
    # 随机抽取
    selected = random.choice(available)
    match_map[idx] = selected
    used_volunteers.add(selected)

# 匹配结果写入新列
df['random_match'] = df.index.map(match_map)

输出示例 #

大数据量可选方案（基于networkx二分图匹配） #

如果数据量较大，可使用二分图匹配库实现更稳定的分配，先安装依赖：
pip install networkx
实现代码：

import networkx as nx

# 构建二分图
G = nx.Graph()
school_nodes = df.index.tolist()
all_volunteers = list({v for lst in df['volunteers'] for v in lst})
G.add_nodes_from(school_nodes, bipartite=0)
G.add_nodes_from(all_volunteers, bipartite=1)

# 建立学校和可选志愿者的关联边
for idx, vol_list in df['volunteers'].items():
    for v in vol_list:
        G.add_edge(idx, v)

# 打乱边顺序实现随机匹配
edges = list(G.edges())
random.shuffle(edges)
G = nx.Graph(edges)

# 计算最大匹配
match_res = nx.bipartite.maximum_matching(G, top_nodes=school_nodes)
df['random_match'] = df.index.map(lambda x: match_res[x])

内容的提问来源于stack exchange，提问作者AbdA