以下代码查找给定图'G'的社区,并根据其所属的社区,在该图内的节点从0-n分配值。然后,代码为每个社区创建新的子图,并在每个社区中找到度数最高的节点。最后,每个子图的顶部节点都集成到整个字典中:
G = 'max : john', 'max : tom', 'jim : john'....'jack : james'
node_partition = dict(community_louvain.best_partition(G))
print node_partition = max: 1, john: 0, james: 3, jim: 4,...tom: 0
"""number of communities = n = list(set(node_partition.values()))"""
dict0 = {k: v for k, v in node_partition.items() if v !=[0]}
G0 = G.copy()
G0.remove_nodes_from(dict0)
degree0 = dict(G.degree(G0))
degree0_dict = dict(sorted(degree0.items(), key=operator.itemgetter(1), reverse=True)[:1])
star_dict = {**degree0_dict, **degree1_dict....**degreek_dict)
此方法有效,但是一个图形可以有n个社区,并且如您所见,以上代码仅适用于社区0中的节点。我必须手动读取已确定的社区数量,并手动重复和编辑每个数字的代码。如何应用自动重复此代码的函数,所以我可以拥有“ n”而不是“ 0”?
答案 0 :(得分:3)
假设您的分区存储在node_partition
中,然后我们制作一个由反向键,值对node_partition
组成的新字典,这将有助于我们以后减少计算复杂性。 (请参阅this for inverting dicitonary和this for getting key with max value in dictionary。)
def invert(d):
"""Turn {a:x, b:x} into {x:[a,b]}"""
r = {}
for k, v in d.items():
r.setdefault(v, []).append(k)
return r
invert_partition = invert(node_partition)
# { 0 :[tom, john] , 1: [mike, elton] ... }
max_deg_per_comm = {}
#iterate over each community
for community_id in invert_partition.keys():
#Extract the sub graph containing the community nodes
temp_graph = G.subgraph(invert_partition[community_id])
#Extract the degrees in the subgraph
temp_degree = dict(temp_graph.degree())
#Store it in a dictionary, with key as community_id and value as the node with max degree
max_deg_per_comm[community_id] = max(temp_degree, key=lambda x: temp_degree[x])
现在,您可以使用字典max_deg_per_comm
获取节点,假设您要查找社区0的节点,请使用
max_deg_per_comm[0]