def plot_embeddings(embeddings, ):
"""
将实体的向量降维到2维，然后显示出来
:param embeddings:
:return:
"""
X, Y = read_node_label(
'/Users/admin/Desktop/GraphEmbedding-deeplearning/data/XunYiWenYao/寻医问药category.txt'
)
emb_list = []
for k in X:
emb_list.append(embeddings[k])
emb_list = np.array(emb_list)
model = TSNE(n_components=2)
node_pos = model.fit_transform(emb_list)
color_idx = {}
for i in range(len(X)):
color_idx.setdefault(Y[i][0], [])
color_idx[Y[i][0]].append(i)
for c, idx in color_idx.items():
plt.scatter(node_pos[idx, 0], node_pos[idx, 1], label=c)
plt.legend()
plt.show()
def plot(embeddings):
emb_list = []
for i in range(1, len(embeddings) + 1):
# print(i)
emb_list.append(embeddings[str(i)])
emb_list = np.array(emb_list)
model = TSNE(n_components=2)
node_pos = model.fit_transform(emb_list)
# print(node_pos)
X, Y = read_node_label('../data/Net300/community.dat', is_net=True)
# colors = ['c', 'b', 'g', 'r', 'm', 'y', 'k']
colors = ['r', 'b', 'g', 'y']
#
color_idx = {}
for i in range(len(X)):
color_idx.setdefault(Y[i][0], [])
color_idx[Y[i][0]].append(i)
# plt.figure(dpi=300, figsize=(24, 12))
for c, idx in color_idx.items():
# print(type(idx))
# print(c,idx)
# print(node_pos[idx, 0],node_pos[idx, 1],node_pos[idx,2])
plt.scatter(node_pos[idx, 0], node_pos[idx, 1], label=c)
# ax.scatter(node_pos[idx, 0], node_pos[idx, 1], node_pos[idx, 2], label=c)
plt.legend() # 图例
plt.show()
def plot_embeddings(embeddings, ):
X, Y = read_node_label('../data/flight/labels-brazil-airports.txt',
skip_head=True)
emb_list = []
for k in X:
emb_list.append(embeddings[k])
emb_list = np.array(emb_list)
model = TSNE(n_components=2)
node_pos = model.fit_transform(emb_list)
color_idx = {}
for i in range(len(X)):
color_idx.setdefault(Y[i][0], [])
color_idx[Y[i][0]].append(i)
for c, idx in color_idx.items():
plt.scatter(node_pos[idx, 0], node_pos[idx, 1],
label=c) # c=node_colors)
plt.legend()
plt.show()
def evaluate_embeddings(embeddings):
X, Y = read_node_label('../data/flight/labels-brazil-airports.txt',
skip_head=True)
tr_frac = 0.8
print("Training classifier using {:.2f}% nodes...".format(tr_frac * 100))
clf = Classifier(embeddings=embeddings, clf=LogisticRegression())
clf.split_train_evaluate(X, Y, tr_frac)
def evaluate_embeddings(embeddings):
X, Y = read_node_label('../data/wiki/wiki_labels.txt')
tr_frac = 0.8
print("Training classifier using {:.2f}% nodes...".format(
tr_frac * 100))
clf = Classifier(embeddings=embeddings, clf=LogisticRegression())
clf.split_train_evaluate(X, Y, tr_frac)
def plot_embeddings_3D(embeddings,):
"""将嵌入降维到3维并可视化
:param embeddings:
"""
X, Y = read_node_label('../data/wiki/wiki_labels.txt')
# print(Y)
emb_list = []
for k in X:
emb_list.append(embeddings[k])
emb_list = np.array(emb_list)
model = TSNE(n_components=3)
node_pos = model.fit_transform(emb_list) # 返回矩阵
# print(node_pos)
color_idx = {}
for i in range(len(X)):
color_idx.setdefault(Y[i][0], [])
color_idx[Y[i][0]].append(i)
# print(color_idx)
# 生产标签_节点字典
fig = plt.figure()
ax = Axes3D(fig)
for c, idx in color_idx.items():
# print(c,idx)
# print(node_pos[idx, 0],node_pos[idx, 1],node_pos[idx,2])
ax.scatter(node_pos[idx, 0], node_pos[idx, 1], node_pos[idx, 2], label=c)
plt.legend() # 图例
plt.show()
def plot_embeddings(embeddings,):
X, Y = read_node_label('../data/wiki/karate.txt')
emb_list = []
embeddings = field_to_json(embeddings)
key = embeddings.keys()
for k in key:
emb_list.append(embeddings[k])
emb_list = np.array(emb_list)
# model = TSNE(n_components=2)
# node_pos = model.fit_transform(emb_list)
# color_idx = {}
# for i in range(len(X)):
# color_idx.setdefault(Y[i][0], [])
# color_idx[Y[i][0]].append(i)
#
# for c, idx in color_idx.items():
# plt.scatter(node_pos[idx, 0], node_pos[idx, 1], label=c)
for i in range(len(key)):
plt.scatter(emb_list[i, 0], emb_list[i, 1], alpha=0.5, s=150, color='b')
plt.text(emb_list[i, 0], emb_list[i, 1], list(key)[i], horizontalalignment='center', verticalalignment='center')
# for j in range(len(X)):
# if X[j] == str(i+1):
# x = int(X[j])
# y = int(Y[j][0])
# plt.plot(node_pos[int(list(key)[x-1])-1], node_pos[int(list(key)[y-1])-1], color='r')
plt.legend()
plt.show()
def evaluate_embeddings(embeddings):
#读入真实的分类label
X, Y = read_node_label('../data/wiki/wiki_labels.txt')
tr_frac = 0.8 #80%的节点用于训练分类器，其余的用于测试
print("Training classifier using {:.2f}% nodes...".format(tr_frac * 100))
#应用分类器对节点进行分类以评估向量的质量
clf = Classifier(embeddings=embeddings, clf=LogisticRegression())
clf.split_train_evaluate(X, Y, tr_frac)
def plot_embeddings(embeddings,):
X, Y = read_node_label('./data/miserables/miserables_labels.txt')
emb_list = []
for k in X:
emb_list.append(embeddings[k])
emb_list = np.array(emb_list)
model = TSNE(n_components=2)
node_pos = model.fit_transform(emb_list)
color_idx = {}
color_dict = {}
for i in range(len(X)):
color_idx.setdefault(Y[i][0], [])
color_idx[Y[i][0]].append(i)
color_dict[X[i]] = COLOR_MAP[int(Y[i][0])]
plt.figure(figsize=(10, 10))
ax = plt.axes([0.05, 0.05, 1 - 0.05 * 2, 1 - 0.05 * 2])
ax.spines["top"].set_visible(False)
ax.spines["right"].set_visible(False)
ax.spines["bottom"].set_visible(False)
ax.spines["left"].set_visible(False)
pos = {}
for i, x in enumerate(X):
pos[x] = node_pos[i]
color_list = []
cluster_color_list = []
for node in G.nodes:
color_list.append(color_dict[node])
nx.draw_networkx_nodes(G, pos, node_size=80, node_color=color_list, edgecolors='white', linewidths=0.7)
#for c, idx in color_idx.items():
#plt.scatter(node_pos[idx, 0], node_pos[idx, 1], label=c) # c=node_colors)
# plt.legend()
t = datetime.datetime.now()
plt.savefig("./pics/miserables/{}-{}-{}/dw-{}-{}-{}-emb-real.png".format(
t.year, t.month, t.day,
t.hour, t.minute, int(t.second)
))
plt.cla()
cluster = kmeans(embeddings, K=8)
cluster_color = []
for node in G.nodes:
cluster_color.append(COLOR_MAP[cluster[node]])
nx.draw_networkx_nodes(G, pos, node_size=80, node_color=cluster_color, edgecolors='white', linewidths=0.7)
plt.savefig("./pics/miserables/{}-{}-{}/dw-{}-{}-{}-emb-cluster.png".format(
t.year, t.month, t.day,
t.hour, t.minute, int(t.second)
))
plt.cla()
return color_dict, cluster
def evaluate_embeddings(embeddings):
X, Y = read_node_label(
'../data/ETH/Phishing node classification/label.txt', skip_head=True)
tr_frac = 0.8
print("Training classifier using {:.2f}% nodes...".format(tr_frac * 100))
clf = Classifier(embeddings=embeddings, clf=LogisticRegression())
clf.split_train_evaluate(X, Y, tr_frac)
def evaluate_embeddings(embeddings):
"""分割训练评估，输出性能参数
:param embeddings:
"""
# X, Y = read_node_label('../data/wiki/wiki_labels.txt')
# X, Y = read_node_label('../data/flight/labels-brazil-airports.txt', True)
# X, Y = read_node_label('../data/flight/labels-europe-airports.txt', True)
X, Y = read_node_label('../data/flight/labels-usa-airports.txt', True)
tr_frac = 0.8 # 交叉验证百分比
print("Training classifier using {:.2%} nodes...".format(tr_frac))
clf = Classifier(embeddings=embeddings, clf=LogisticRegression())
return clf.split_train_evaluate(X, Y, tr_frac)
def evaluate_embeddings(embeddings):
"""
一个分类器函数，用来评价向量好坏，因为每个实体有对应的标签，通过向量实现多分类
:param embeddings:
:return:
"""
X, Y = read_node_label(
'/Users/admin/Desktop/GraphEmbedding-deeplearning/data/XunYiWenYao/寻医问药category.txt'
)
tr_frac = 0.8
print("Training classifier using {:.2f}% nodes...".format(tr_frac * 100))
clf = Classifier(embeddings=embeddings, clf=LogisticRegression())
clf.split_train_evaluate(X, Y, tr_frac)
def NMI_Q(embeddings, num_coms):
emb_list = []
for i in range(1, len(embeddings) + 1):
emb_list.append(embeddings[str(i)])
# edgelist index start at 1 / emb_list index start at 0
predict = kmeans_from_vec(emb_list, num_coms) # index start by 0
for i in range(len(predict)):
predict[i] = [str(x) for x in predict[i]]
# 数据处理
X, Y = read_node_label('../data/Net_mu/Net0.8/community.dat', is_net=True)
comu_idx = {}
for i in range(len(X)):
comu_idx.setdefault(Y[i][0], [])
comu_idx[Y[i][0]].append(i)
# print(comu_idx)
real = []
# print(real)
for key in comu_idx:
real.append(comu_idx[key])
for i in range(len(real)):
real[i] = [str(x) for x in real[i]]
# print(predict)
# print(real)
mni = NMI(predict, real)
# print(mni)
#
predict_ = predict
for i in range(len(predict)):
predict_[i] = [str(int(x) + 1) for x in predict[i]]
# predict index add 1
#
#
q = Q(predict_, G)
# print(q)
#
return mni, q
def plot_embeddings(embeddings, ):
X, Y = read_node_label('../data/wiki/karate.txt', skip_head=True)
emb_list = []
for k in X:
emb_list.append(embeddings[k])
emb_list = np.array(emb_list)
# model = TSNE(n_components=2)
# node_pos = model.fit_transform(emb_list)
# color_idx = {}
# for i in range(len(X)):
# color_idx.setdefault(Y[i][0], [])
# color_idx[Y[i][0]].append(i)
for idx in range(len(X)):
plt.scatter(emb_list[idx, 0], emb_list[idx, 1])
plt.legend()
plt.show()
def evaluate_embeddings(embeddings):
X, Y = read_node_label('../data/flight/igraph2Label.txt', skip_head=True)
# tr_frac = 0.8
#
# print("Training classifier using {:.2f}% nodes...".format(
#
# tr_frac * 100))
#
# # clf = Classifier(embeddings=embeddings, clf=LogisticRegression())
# #
# # clf.split_train_evaluate(X, Y, tr_frac)
# X_ = numpy.asarray([embeddings[x] for x in X])
# print("kexinxin")
X = numpy.asarray([embeddings[x] for x in X])
x_train, x_test, y_train, y_test = train_test_split(X, Y, test_size=0.4)
clf = linear_model.LogisticRegression(solver='liblinear')
# clf = RandomForestClassifier()
# clf=tree.DecisionTreeClassifier(criterion='entropy')
#clf = KNeighborsClassifier()
clf.fit(x_train, y_train)
y_pred = clf.predict(x_test)
# count=0
# # for i in range(len(result)):
# # if result[i]==y_test[i]:
# # count=count+1
# # print(count/len(result))
y_test = np.array(y_test)
y_test = y_test.astype(np.int16)
y_pred = y_pred.astype(np.int16)
#y_pred.reshape(-1,1)
#y_test.reshape(-1,1)
#y_pred=y_pred.tolist()
#arr=np.array([1,2,3,4])
print("acc", accuracy_score(y_test, y_pred))
#print("precision", precision_score(y_test, y_pred))
print("recall", recall_score(y_test, y_pred))
print("f1", f1_score(y_test, y_pred))
def plot_embeddings(embeddings, ):
X, Y = read_node_label('../data/wiki/friend_labels_baseline.txt')
emb_list = []
for k in X:
emb_list.append(embeddings[k])
emb_list = np.array(emb_list)
model = TSNE(n_components=2)
node_pos = model.fit_transform(emb_list)
color_idx = {}
for i in range(len(X)):
color_idx.setdefault(Y[i][0], [])
color_idx[Y[i][0]].append(i)
for c, idx in color_idx.items():
plt.scatter(node_pos[idx, 0], node_pos[idx, 1], label=c)
plt.legend()
plt.show()
def plot_embeddings(embeddings,):
X, Y = read_node_label('../data/wiki/wiki_labels.txt')
emb_list = []
for k in X:
emb_list.append(embeddings[k])
emb_list = np.array(emb_list)
model = TSNE(n_components=2)
node_pos = model.fit_transform(emb_list)
color_idx = {}
for i in range(len(X)):
color_idx.setdefault(Y[i][0], [])
color_idx[Y[i][0]].append(i)
# 分辨率参数-dpi，画布大小参数-figsize
plt.figure(dpi=300, figsize=(24, 12))
for c, idx in color_idx.items():
plt.scatter(node_pos[idx, 0], node_pos[idx, 1], label=c)
plt.legend()
plt.show()
def plot(embeddings):
emb_list = []
for i in range(1, len(embeddings) + 1):
emb_list.append(embeddings[str(i)])
# 将嵌入向量字典转换成列表
emb_list = np.array(emb_list)
model = TSNE(n_components=2)
node_pos = model.fit_transform(emb_list)
# 通过tsne转换成二维坐标
X, Y = read_node_label('../data/Net300/community.dat', is_net=True)
# ['1', '2', '3', '4'],[['4'], ['2'], ['4'], ['1']]
commu_idx = {}
for i in range(len(X)):
commu_idx.setdefault(Y[i][0], [])
commu_idx[Y[i][0]].append(i)
for c, idx in commu_idx.items():
plt.scatter(node_pos[idx, 0], node_pos[idx, 1], label=c)
plt.legend()
plt.show()
def plot_embeddings(embeddings,):
"""将嵌入降维到2维并可视化
:param embeddings:
"""
X, Y = read_node_label('../data/wiki/wiki_labels.txt')
# X, Y = read_node_label('../data/flight/labels-brazil-airports.txt', True)
# X, Y = read_node_label('../data/flight/labels-europe-airports.txt', True)
# X, Y = read_node_label('../data/flight/labels-usa-airports.txt', True)
# print(Y)
emb_list = []
for k in X:
emb_list.append(embeddings[k])
emb_list = np.array(emb_list)
model = TSNE(n_components=2)
# model = TSNE(n_components=3)
node_pos = model.fit_transform(emb_list) # 返回矩阵
# print(node_pos)
color_idx = {}
for i in range(len(X)):
color_idx.setdefault(Y[i][0], [])
color_idx[Y[i][0]].append(i)
# print(color_idx)
# 生产标签_节点字典
# fig = plt.figure()
# ax = Axes3D(fig)
# 分辨率参数-dpi，画布大小参数-figsize
plt.figure(dpi=300, figsize=(24, 12))
for c, idx in color_idx.items():
# print(type(idx))
# print(c,idx)
# print(node_pos[idx, 0],node_pos[idx, 1],node_pos[idx,2])
plt.scatter(node_pos[idx, 0], node_pos[idx, 1], label=c)
# ax.scatter(node_pos[idx, 0], node_pos[idx, 1], node_pos[idx, 2], label=c)
# plt.legend() # 图例
plt.show()
plt.scatter(node_pos[idx, 0], node_pos[idx, 1],
label=c) # c=node_colors)
plt.legend()
plt.show()
d_set = [
"Facebook-Page2Page", "PubMed-Diabetes", "Terrorists-Relation"
] # ["Cora", "CiteSeer", "Facebook-Page2Page", "PubMed-Diabetes", "Terrorists-Relation", "Zachary-Karate", "Internet-Industry-Partnerships"] # [sparse, dense]
mdl = ["Node2Vec", "SDNE", "DeepWalk",
"LINE"] # ["Node2Vec", "SDNE", "Struc2Vec", "DeepWalk", "LINE"]
for i in range(len(d_set)):
# Load/Prepare data
graph_fname = "data/" + d_set[i] + "/" + d_set[i]
X, Y = read_node_label(graph_fname + ".labels", skip_head=True)
X = np.asarray(X)
Y = np.asarray(Y)
# Preserve ratio/percentage of samples per class using efficent data-splitting && data-resampling strageies
train_frac = 0.8
test_frac = round((1 - train_frac), 1)
print("Training classifier using {:.2f}% nodes...".format(train_frac *
100))
if not os.path.isfile(graph_fname + "_strat_train_test.splits"):
stratified_data = StratifiedShuffleSplit(n_splits=1,
test_size=test_frac,
train_size=train_frac,
random_state=42)
for train_index, test_index in stratified_data.split(X, Y):
strat_X_train, strat_y_train = X[train_index], Y[train_index]
