美国大学生足球联赛数据集football——node2vec

# 使用Node2Vec对space_data进行压缩
import networkx as nx
from tqdm import tqdm
from sklearn.decomposition import PCA
from node2vec import Node2Vec
import matplotlib.pyplot as plt
import warnings
warnings.filterwarnings('ignore')

# 数据加载，构造图
G = nx.read_gml('football.gml')
print(type(G))
#G = nx.read_gml('football.gml', relabel=True)
print(len(G))

# 初始化Node2Vec模型
#model = DeepWalk(G, walk_length=10, num_walks=5, workers=1)
model = Node2Vec(G, walk_length = 10, num_walks = 5, p = 0.25, q = 4, workers = 1)
# 模型训练
result = model.fit(window=4, iter=20)
# 得到节点的embedding
print(result.wv.most_similar('EastCarolina'))
embeddings = result.wv
print(embeddings)

# 在二维空间中绘制所选节点的向量
def plot_nodes(word_list):
    # 每个节点的embedding为100维
    X = []
    for item in word_list:
        X.append(embeddings[item])
    #print(X.shape)
    # 将100维向量减少到2维
    pca = PCA(n_components=2)
    result = pca.fit_transform(X) 
    #print(result)
    # 绘制节点向量
    plt.figure(figsize=(12,9))
    # 创建一个散点图的投影
    plt.scatter(result[:, 0], result[:, 1])
    for i, word in enumerate(list(word_list)):
        plt.annotate(word, xy=(result[i, 0], result[i, 1]))        
    plt.show()
    
plot_nodes(result.wv.vocab)