在R语言的iGraph中识别有向网络中的双向关注匹配

Hey Kimberly, let's break this down step by step for your Twitter directed graph analysis using iGraph—this is a super common use case, so we've got solid, actionable patterns to follow.

1. 识别双向关注（互粉）关系 #

In a directed graph, mutual follows mean there's both an edge from User A to User B and from User B to User A. iGraph has built-in tools to streamline this, but we can also make explicit checks for clarity:

import igraph as ig

# 假设你已经构建好有向图，命名为g
# 方法1：用iGraph内置的互边标记
mutual_edges = g.es.select(_is_mutual=True)
# 提取去重的互粉节点对（避免重复记录(A,B)和(B,A)）
mutual_node_pairs = [(e.source, e.target) for e in mutual_edges if e.source < e.target]

# 方法2：手动验证反向边（适合理解逻辑）
mutual_pairs = set()
for edge in g.es:
    source, target = edge.source, edge.target
    if g.are_connected(target, source):
        # 用sorted确保每对只存一次
        pair = tuple(sorted((source, target)))
        mutual_pairs.add(pair)

2. 筛选未获得回关的用户 #

We can split this into two groups: users who follow others but aren't followed back, and users who are followed but don't follow back. Here's how to extract both:

# 获取所有单向关注的边（A→B存在，但B→A不存在）
unidirectional_edges = g.es.select(_is_mutual=False)

# 统计两类用户
unfollowed_followers = set()  # 关注了别人却没被回关的用户
unfollowing_targets = set()   # 被别人关注却没回关的用户

for edge in unidirectional_edges:
    a, b = edge.source, edge.target
    unfollowed_followers.add(a)
    unfollowing_targets.add(b)

# 关联你的用户性别数据（假设user_data是{节点ID: {'gender': 'male/female/other'}}）
unfollowed_male_count = len([uid for uid in unfollowed_followers if user_data[uid]['gender'] == 'male'])
unfollowed_female_count = len([uid for uid in unfollowed_followers if user_data[uid]['gender'] == 'female'])

3. 分析性别与回关率的关联 #

To test if male users get more follow-backs, we'll calculate follow-back rates (mutual follows ÷ total followers) for each gender, then run basic stats to check for significance:

import numpy as np
from scipy.stats import ttest_ind

# 计算每个用户的回关率
followback_rates = {}
for node in g.vs:
    node_id = node.index
    total_followers = node.indegree()
    if total_followers == 0:
        continue  # 跳过无粉丝的用户
    
    # 统计该用户回关了多少粉丝
    mutual_count = 0
    for follower in g.neighbors(node_id, mode='in'):
        if g.are_connected(node_id, follower):
            mutual_count += 1
    
    followback_rates[node_id] = mutual_count / total_followers

# 按性别分组计算平均回关率
male_rates = [rate for uid, rate in followback_rates.items() if user_data[uid]['gender'] == 'male']
female_rates = [rate for uid, rate in followback_rates.items() if user_data[uid]['gender'] == 'female']

male_avg = np.mean(male_rates) if male_rates else 0
female_avg = np.mean(female_rates) if female_rates else 0

print(f"男性用户平均回关率: {male_avg:.2f}")
print(f"女性用户平均回关率: {female_avg:.2f}")

# 做显著性检验（验证差异是否偶然）
if len(male_rates) > 1 and len(female_rates) > 1:
    stat, p_value = ttest_ind(male_rates, female_rates)
    print(f"T检验统计量: {stat:.2f}, P值: {p_value:.4f}")
    print("差异具有统计学显著性" if p_value < 0.05 else "差异无统计学显著性")

关键注意事项 #

• 节点ID映射: 确保你的iGraph节点ID和用户数据表的ID完全对应，避免数据错位
• 缺失数据: 单独处理性别未知的用户（可以排除或单独分组，避免干扰结果）
• 性能优化: 对于超大型数据集，用iGraph的get_adjacency()矩阵代替循环遍历，会大幅提升速度

Let me know if you need help tweaking this to fit your specific dataset structure or if you hit performance snags with large graphs!

内容的提问来源于stack exchange，提问作者Kimberly