相关资料

The Louvain method for community detection内含Java代码。

Louvain method是一个非重叠社团发现的算法，该方法具有较快的执行效率。

对应的paper地址：Louvain method: Finding communities in large networks

Paper作者对算法的介绍网址：http://perso.uclouvain.be/vincent.blondel/research/louvain.html

c++版原作者代码：https://sourceforge.net/projects/louvain/
Rabbit: 复现+并行版本

Engineering Parallel Algorithms forCommunity Detection in Massive Networks, 并行版本

Fast parallel algorithm for unfolding of communities in large graphs

博客介绍：

• [社区发现相关算法]

• Modularity的计算方法——社团检测中模块度计算公式详解

• 模块度与Louvain社区发现算法

• https://www.cnblogs.com/allanspark/p/4197980.html

• 基于模块度在有向图中进行社区发现
  Community structure in directed networks

代码实现

networkx

 pip3 install python-louvain
 pip3 install networkx

测试：

# basic usage
import community as community_louvain
import networkx as nx

G = nx.erdos_renyi_graph(100, 0.01)
partion = community_louvain.best_partition(G)

print(partion)

result

{0: 0, 1: 1, 2: 2, 3: 3, 4: 4, 5: 5, 6: 3, 7: 6, 8: 7, 9: 8, 10: 9, 11: 10, 12: 11, 13: 12, 14: 6, 15: 5, 16: 1, 17: 31, 18: 14, 19: 3, 20: 15, 21: 16, 22: 17, 23: 10, 24: 6, 25: 31, 26: 18, 27: 0, 28: 19, 29: 5, 30: 20, 31: 21, 32: 20, 33: 20, 34: 22, 35: 23, 36: 24, 37: 25, 38: 26, 39: 3, 40: 27, 41: 15, 42: 3, 43: 20, 44: 1, 45: 1, 46: 28, 47: 5, 48: 29, 49: 30, 50: 6, 51: 20, 52: 32, 53: 7, 54: 1, 55: 20, 56: 5, 57: 31, 58: 33, 59: 31, 60: 15, 61: 6, 62: 34, 63: 15, 64: 35, 65: 24, 66: 31, 67: 36, 68: 37, 69: 3, 70: 6, 71: 1, 72: 15, 73: 38, 74: 39, 75: 37, 76: 20, 77: 1, 78: 5, 79: 31, 80: 31, 81: 6, 82: 36, 83: 31, 84: 41, 85: 42, 86: 3, 87: 43, 88: 15, 89: 31, 90: 20, 91: 6, 92: 5, 93: 3, 94: 40, 95: 13, 96: 6, 97: 20, 98: 4, 99: 31}

测试真实数据：

import networkx as nx
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import datetime

edgestring = ''
path = 'test_weighted_0.e'
# path = 'dataset/web3.txt'
with open(path,  'r+', encoding='utf-8') as f:
    edgestring = f.read()

edgeList = edgestring.split('\n')
G = nx.DiGraph()

print('加入边集...')
nodes_set = set()
for edge in edgeList:
    if len(edge) <= 0:
        continue
    edge = edge.split(' ')
    G.add_edge(int(edge[0]) , int(edge[1]) )   #一次添加一条边
    print(edge[0], edge[1])

# 绘图
# nx.draw(G, with_labels=True , node_size = 200)
# plt.show()

start = datetime.datetime.now()
pr = nx.pagerank(G, alpha=0.85, personalization=None,
             max_iter=100, tol=1.0e-5, nstart=None, 
             dangling=None)
end =  datetime.datetime.now()
for i in range(0, len(G.nodes)): 
    print(i, pr[i])

print('用时:', end-start)

# print('G:', G.edges)
# print('G:', G.nodes)

Directed Louvain : maximizing modularity in directed networks

paper
code