结果对不上，大佬 about communitydetection HOT 9 OPEN

我用了club数据集，跑出来的Q值是0.2869而不是CSDN展示的0.38，请问是什么问题呢？

我这里是正常的，可能是随机数的影响？

from communitydetection.

G = load_graph('club.txt')
obj = Graph()
G1 = obj.createGraph("club.txt")

start_time = time.time()
algorithm = Louvain(G)
communities = algorithm.execute()
end_time = time.time()
# 按照社区大小从大到小排序输出
communities = sorted(communities, key=lambda b: -len(b))  # 按社区大小排序
count = 0
for communitie in communities:
    count += 1
    print("社区", count, " ", communitie)

print(cal_Q(communities, G1))
print(f'算法执行时间{end_time - start_time}')

from communitydetection.

G = load_graph('club.txt')
obj = Graph()
G1 = obj.createGraph("club.txt")

start_time = time.time()
algorithm = Louvain(G)
communities = algorithm.execute()
end_time = time.time()
# 按照社区大小从大到小排序输出
communities = sorted(communities, key=lambda b: -len(b))  # 按社区大小排序
count = 0
for communitie in communities:
    count += 1
    print("社区", count, " ", communitie)

print(cal_Q(communities, G1))
print(f'算法执行时间{end_time - start_time}')

主要代码是这样的，应该也没什么能错的地方啊。。。

from communitydetection.

G = load_graph('club.txt')
obj = Graph()
G1 = obj.createGraph("club.txt")

start_time = time.time()
algorithm = Louvain(G)
communities = algorithm.execute()
end_time = time.time()
# 按照社区大小从大到小排序输出
communities = sorted(communities, key=lambda b: -len(b))  # 按社区大小排序
count = 0
for communitie in communities:
    count += 1
    print("社区", count, " ", communitie)

print(cal_Q(communities, G1))
print(f'算法执行时间{end_time - start_time}')

主要代码是这样的，应该也没什么能错的地方啊。。。

最终分区的结果是一样的，就是Q值不对。

from communitydetection.

G = load_graph('club.txt')
obj = Graph()
G1 = obj.createGraph("club.txt")

start_time = time.time()
algorithm = Louvain(G)
communities = algorithm.execute()
end_time = time.time()
# 按照社区大小从大到小排序输出
communities = sorted(communities, key=lambda b: -len(b))  # 按社区大小排序
count = 0
for communitie in communities:
    count += 1
    print("社区", count, " ", communitie)

print(cal_Q(communities, G1))
print(f'算法执行时间{end_time - start_time}')

主要代码是这样的，应该也没什么能错的地方啊。。。

最终分区的结果是一样的，就是Q值不对。

obj.createGraph有点问题，邻居没重复计算。所以我用的G1 = nx.karate_club_graph()。以后有空修改下。

from communitydetection.

改为无向图就行了， self.graph = nx.Graph()

from communitydetection.

改为无向图就行了， self.graph = nx.Graph()

哇，确实解决了，感谢大佬。

from communitydetection.

我用了俱乐部数据集,跑出来的Q值是0.2869而不是CSDN展示的0.38,请问是什么问题呢？
有偿求教

from communitydetection.

想求教这个louvain算法的整个流程讲解，我需要对其进行改进，有偿求教，微信号LBJgigigi13

from communitydetection.