GCN实现节点分类任务

数据处理，数据集采用cora数据集

import dgl
from dgl.data import DGLDataset
import torch
import os
import pandas as pd
import torch
import numpy as np
import matplotlib.pyplot as plt
import networkx as nx
import os
os.environ["KMP_DUPLICATE_LIB_OK"]="TRUE"
import torch.nn.functional as F

class CoraDataset(DGLDataset):
    def __init__(self):
        super().__init__(name='cora')

    def process(self):
        nodes_data = pd.read_csv(r'../data/cora/cora.content', sep='\t', header=None)
        edges_data = pd.read_csv(r'../data/cora/cora.cites', sep='\t', header=None)
        node_f = nodes_data.iloc[:, 1:-1]
        node_len = nodes_data.iloc[:, 0]
        node_l = nodes_data.iloc[:, -1]
        edges_s = edges_data.iloc[:, 0]
        edges_d = edges_data.iloc[:, 1]
        idx = np.array(node_len, dtype=np.int32)

        idx_map = {j: i for i, j in enumerate(idx)}
        #print(idx_map)

        node_features = torch.from_numpy(node_f.to_numpy())

        node_labels = torch.from_numpy(node_l.astype('category').cat.codes.to_numpy().astype(np.int64))
        edges_src = torch.from_numpy(np.array(list(map(idx_map.get, edges_s))))
        edges_dst = torch.from_numpy(np.array(list(map(idx_map.get, edges_d))))



        #print(edges_src, edges_dst)
        self.graph = dgl.graph((edges_src, edges_dst), num_nodes=nodes_data.shape[0])
        self.graph.ndata['feat'] = node_features
        self.graph.ndata['label'] = node_labels

        n_nodes = nodes_data.shape[0]
        n_train = int(n_nodes * 0.6)
        n_val = int(n_nodes * 0.2)
        train_mask = torch.zeros(n_nodes, dtype=torch.bool)
        val_mask = torch.zeros(n_nodes, dtype=torch.bool)
        test_mask = torch.zeros(n_nodes, dtype=torch.bool)
        train_mask[:n_train] = True
        val_mask[n_train:n_train + n_val] = True
        test_mask[n_train + n_val:] = True
        self.graph.ndata['train_mask'] = train_mask
        self.graph.ndata['val_mask'] = val_mask
        self.graph.ndata['test_mask'] = test_mask

    def __getitem__(self, i):
        return self.graph

    def __len__(self):
        return 1

数据 导入neo4j中查看是否存在孤立节点

#查找入度为0的节点
match (n) where not()-[]-(n) return n

查找出度为0的节点
match (n) where not(n)-[]-() return n

模型搭建及训练

from dgl.nn import GraphConv
import torch.nn as nn
import torch.nn.functional as F
import torch
import dgl
from DLG_GCN.dataprocess import CoraDataset
import matplotlib.pyplot as plt
import numpy as np


class GCN(nn.Module):
    def __init__(self, in_feats, h_feats, num_classes):
        super(GCN, self).__init__()
        self.conv1 = GraphConv(in_feats, h_feats)
        self.conv2 = GraphConv(h_feats, num_classes)

    def forward(self, g, in_feat):
        h = self.conv1(g, in_feat)
        h = F.relu(h)
        h = self.conv2(g, h)
        return h




def train(g, model):
    optimizer = torch.optim.Adam(model.parameters(), lr = 0.0005)
    best_val_acc = 0
    best_test_acc = 0

    fig_loss = []
    fig_test_acc = []
    fig_val_acc = []

    features = g.ndata['feat']

    labels = g.ndata['label']
    train_mask = g.ndata['train_mask']
    val_mask = g.ndata['val_mask']
    test_mask = g.ndata['test_mask']

    for epoch in range(1000):
        logits = model(g, features)
        pred = logits.argmax(1)

        loss = F.cross_entropy(logits[train_mask], labels[train_mask])

        # Compute accuracy on training/validation/test
        train_acc = (pred[train_mask] == labels[train_mask]).float().mean()
        val_acc = (pred[val_mask] == labels[val_mask]).float().mean()
        test_acc = (pred[test_mask] == labels[test_mask]).float().mean()

        fig_loss.append(loss)
        fig_val_acc.append(val_acc)
        fig_test_acc.append(test_acc)

        if best_val_acc < val_acc:
            best_val_acc = val_acc
            best_test_acc = test_acc

        # Backward
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        if epoch % 5 == 0:
            print('In epoch {}, loss: {:.3f}, val acc: {:.3f} (best {:.3f}), test acc: {:.3f} (best {:.3f})'.format(
                epoch, loss, val_acc, best_val_acc, test_acc, best_test_acc))
    fig, ax1 = plt.subplots()
    ax2 = ax1.twinx()
    ln1 = plt.plot(np.arange(len(fig_loss)), fig_loss, 'r', label='loss')
    ln2 = plt.plot(np.arange(len(fig_test_acc)), fig_test_acc, 'g', label='test_acc')
    ln3 = plt.plot(np.arange(len(fig_val_acc)), fig_val_acc, 'b', label='val_acc')
    ax1.set_xlabel('iteration')
    ax2.set_ylabel('training loss')
    ax2.set_ylabel('training accuracy')

    lns = ln1 + ln2 +ln3
    labels = ["train_loss", "test_acc", "val_acc"]
    # labels = [l.get_label() for l in lns]
    plt.legend(lns, labels, loc='upper left')
    plt.grid(True)

    plt.savefig("./mydata/train_val_test.png")

    plt.figure(2)
    plt.plot(np.arange(len(fig_loss)), fig_loss, 'r', label='loss')
    plt.legend(labels=['loss'])
    plt.grid(True)
    plt.savefig("./mydata/train.png")

    plt.figure(3)
    plt.plot(np.arange(len(fig_val_acc)), fig_val_acc, 'g', label='val_accuracy')
    plt.legend(labels=["test_acuracy"])
    plt.grid(True)
    plt.savefig("./mydata/val.png")

    plt.figure(4)
    plt.plot(np.arange(len(fig_test_acc)), fig_test_acc, 'b', label='test_accuracy')
    plt.legend(labels=["test_acuracy"])
    plt.grid(True)
    plt.savefig("./mydata/test.png")

    plt.show()

cora_data = CoraDataset()
graph = cora_data[0]

g = dgl.to_bidirected(graph, copy_ndata=True)

model = GCN(g.ndata['feat'].shape[1], 16, 7)

train(g, model)

训练结果