融合全局信息和对比学习的图神经网络推荐模型
王彦乐,张瑞峰,李锵

Graph neural network recommendation model integrating global information and contrastive learning
Yanle WANG,Ruifeng ZHANG,Qiang LI
表 2 不同推荐模型在4个公开数据集上的性能参数对比
Tab.2 Performance comparison of different recommendation models on four public datasets
数据集模型Recall@10NDCG@10Recall@20NDCG@20Recall@50NDCG@50
ML-1MNGCF0.184 60.252 80.274 10.261 40.434 10.305 5
LightGCN0.187 60.251 40.279 60.262 00.446 90.309 1
SGL0.188 80.252 60.284 80.264 90.448 70.311 1
SimGCL0.205 20.272 60.298 80.281 50.451 50.328 2
NCL0.204 80.272 20.302 50.283 20.462 80.329 2
AdaGCL0.207 20.274 50.301 20.286 80.466 20.329 5
GraphAug0.206 00.273 20.303 40.284 30.468 60.330 0
GICL0.210 30.277 30.304 40.287 90.470 60.333 9
η/%+1.50+1.02+0.33+0.38+0.43+1.18
YelpNGCF0.063 00.044 60.102 60.056 70.186 40.078 4
LightGCN0.073 00.052 00.116 30.065 20.201 60.087 5
SGL0.083 30.060 10.128 80.073 90.214 00.096 4
SimGCL0.089 60.065 80.130 40.081 00.209 80.101 8
NCL0.091 20.066 20.132 60.081 30.224 70.104 8
AdaGCL0.090 20.067 30.131 20.082 20.221 30.102 5
GraphAug0.092 20.068 20.137 70.081 70.224 00.104 2
GICL0.095 80.070 40.143 40.084 20.233 20.107 4
η/%+3.90+3.23+4.14+2.43+3.78+2.48
BooksNGCF0.061 70.042 70.097 80.053 70.169 90.072 5
LightGCN0.079 70.056 50.120 60.068 90.201 20.089 9
SGL0.089 80.064 50.133 10.077 70.215 70.099 2
SimGCL0.092 40.065 60.133 80.082 00.215 80.100 9
NCL0.093 00.066 40.137 70.081 20.216 40.101 7
AdaGCL0.094 20.067 20.134 60.082 80.217 70.102 3
GraphAug0.093 70.068 20.138 10.081 50.217 00.103 2
GICL0.098 90.071 50.146 40.085 90.230 70.108 2
η/%+4.99+4.84+6.01+3.74+5.97+4.84
GowallaNGCF0.119 20.085 20.175 50.101 30.281 10.127 0
LightGCN0.136 20.087 60.197 60.115 20.304 40.141 4
SGL0.146 50.104 80.208 40.122 50.319 70.149 7
SimGCL0.147 20.105 20.200 50.123 20.319 80.150 8
NCL0.148 20.106 10.211 80.125 30.322 90.153 1
AdaGCL0.149 10.108 20.212 70.127 50.323 80.152 2
GraphAug0.151 20.107 30.213 20.126 50.325 40.154 1
GICL0.156 70.113 90.220 50.131 80.332 70.159 6
η/%+3.64+5.27+3.42+3.37+2.24+3.57