KGCN
KGCN copied to clipboard
造不出来同款数据集,求解
Hello Wang, Book-Crossing 的 item_index2entity_id.txt 和kg.txt 可否给个获取途径,邮箱[email protected] THanks
你好,请参考https://github.com/hwwang55/RippleNet/tree/master/data/book。谢谢!
https://github.com/hwwang55/RippleNet/tree/master/data/book:
运行preprocess.py 结果同RippleNet相同,但与KGCN差距有点大,不知需要做什么转换嘛?
KGCN论文中book信息
number of users: 19676 number of items: 20003 number of entities (containing items): 25787 number of relations: 18
实际运行
BX-Book-Ratings.csv kg_rehashed.txt rename kg.txt item_index2entity_id_rehashed.txt rename item_index2entity_id
运行结果
number of users: 17860 number of items: 14967 number of entities (containing items): 77903 number of relations: 25
KGCN main运行结果:
epoch 0 train auc: 0.5299 f1: 0.5128 eval auc: 0.4939 f1: 0.5218 test auc: 0.4962 f1: 0.5325
epoch 1 train auc: 0.6575 f1: 0.6395 eval auc: 0.5702 f1: 0.5959 test auc: 0.5673 f1: 0.6021
epoch 2 train auc: 0.8135 f1: 0.7262 eval auc: 0.6838 f1: 0.6069 test auc: 0.6830 f1: 0.6074
epoch 3 train auc: 0.8637 f1: 0.7820 eval auc: 0.6869 f1: 0.6332 test auc: 0.6859 f1: 0.6349
epoch 4 train auc: 0.8791 f1: 0.7936 eval auc: 0.6855 f1: 0.6365 test auc: 0.6855 f1: 0.6373
epoch 5 train auc: 0.8911 f1: 0.8034 eval auc: 0.6839 f1: 0.6369 test auc: 0.6848 f1: 0.6380
epoch 6 train auc: 0.9025 f1: 0.8169 eval auc: 0.6810 f1: 0.6339 test auc: 0.6822 f1: 0.6378
epoch 7 train auc: 0.9124 f1: 0.8264 eval auc: 0.6790 f1: 0.6302 test auc: 0.6800 f1: 0.6326
epoch 8 train auc: 0.9186 f1: 0.8357 eval auc: 0.6760 f1: 0.6321 test auc: 0.6771 f1: 0.6344
epoch 9 train auc: 0.9237 f1: 0.8410 eval auc: 0.6738 f1: 0.6301 test auc: 0.6752 f1: 0.6319
Thanks@hwwang55
您好,请问如果我需要构造其他数据集,比如movieLens100k或者movieLens-small数据集对应的itemid2entityid.txt和kg.txt具体应该怎么做呀?可否给一种参考方式,联系方式[email protected]谢谢!
您好,请问如果我需要构造其他数据集,比如movieLens100k或者movieLens-small数据集对应的itemid2entityid.txt和kg.txt具体应该怎么做呀?可否给一种参考方式,联系方式[email protected]谢谢!
请问你知道怎么做吗?