Can the authors provide the data preprocess code?

[Weile0409]

Dear author, I like your work very much!
Can you provide the data preprocess code (from raw data to the data shown in your paper)?
Thank you !

[xizhu1022]

Dear author, I like your work very much! Can you provide the data preprocess code (from raw data to the data shown in your paper)? Thank you !

Hi, friend. I have the same the problem, have you had the data preprocess code or an alternative way? Looking for your reply! Thanks!

[Weile0409]

Dear author, I like your work very much! Can you provide the data preprocess code (from raw data to the data shown in your paper)? Thank you !

Hi, friend. I have the same the problem, have you had the data preprocess code or an alternative way? Looking for your reply! Thanks!

Hey bro, not yet :(
I think you can use Taobao and Beibei dataset, which are commonly used in multi-behavior recommendation (You can refer to the paper GHCF, EHCF).

[xizhu1022]

Dear author, I like your work very much! Can you provide the data preprocess code (from raw data to the data shown in your paper)? Thank you !

Hi, friend. I have the same the problem, have you had the data preprocess code or an alternative way? Looking for your reply! Thanks!

Hey bro, not yet :( I think you can use Taobao and Beibei dataset, which are commonly used in multi-behavior recommendation (You can refer to the paper GHCF, EHCF).

Thanks for your kind reply.
Yelp, the train/test mat files could be simply processed with the open-source datasets and the code context.
But you see that the CML model adopted a meta-learning framework that the authors preprocessed a related file which is obscure. In fact, I could not reproduce the experiments without these files or more details. What a pity! @anonymousCML

CML/main.py

Lines 46 to 60 in 00b9a5b

	self.trn_file = args.path + args.dataset + '/trn_'
	self.tst_file = args.path + args.dataset + '/tst_int'
	# self.tst_file = args.path + args.dataset + '/BST_tst_int_59'
	#Tmall: 3,4,5,6,8,59
	#IJCAI_15: 5,6,8,10,13,53
	# self.meta_multi_file = args.path + args.dataset + '/meta_multi_beh_user_index'
	# self.meta_single_file = args.path + args.dataset + '/meta_single_beh_user_index'
	self.meta_multi_single_file = args.path + args.dataset + '/meta_multi_single_beh_user_index_shuffle'
	# /meta_multi_single_beh_user_index_shuffle
	# /new_multi_single
	# self.meta_multi = pickle.load(open(self.meta_multi_file, 'rb'))
	# self.meta_single = pickle.load(open(self.meta_single_file, 'rb'))
	self.meta_multi_single = pickle.load(open(self.meta_multi_single_file, 'rb'))

[Rocket-ZZW]

Dear author, I like your work very much! Can you provide the data preprocess code (from raw data to the data shown in your paper)? Thank you !

Hi, friend. I have the same the problem, have you had the data preprocess code or an alternative way? Looking for your reply! Thanks!

Hey bro, not yet :( I think you can use Taobao and Beibei dataset, which are commonly used in multi-behavior recommendation (You can refer to the paper GHCF, EHCF).

Thanks for your kind reply. Yelp, the train/test mat files could be simply processed with the open-source datasets and the code context. But you see that the CML model adopted a meta-learning framework that the authors preprocessed a related file which is obscure. In fact, I could not reproduce the experiments without these files or more details. What a pity! @anonymousCML

CML/main.py

Lines 46 to 60 in 00b9a5b

	self.trn_file = args.path + args.dataset + '/trn_'
	self.tst_file = args.path + args.dataset + '/tst_int'
	# self.tst_file = args.path + args.dataset + '/BST_tst_int_59'
	#Tmall: 3,4,5,6,8,59
	#IJCAI_15: 5,6,8,10,13,53
	# self.meta_multi_file = args.path + args.dataset + '/meta_multi_beh_user_index'
	# self.meta_single_file = args.path + args.dataset + '/meta_single_beh_user_index'
	self.meta_multi_single_file = args.path + args.dataset + '/meta_multi_single_beh_user_index_shuffle'
	# /meta_multi_single_beh_user_index_shuffle
	# /new_multi_single
	# self.meta_multi = pickle.load(open(self.meta_multi_file, 'rb'))
	# self.meta_single = pickle.load(open(self.meta_single_file, 'rb'))
	self.meta_multi_single = pickle.load(open(self.meta_multi_single_file, 'rb'))

Hi, friend, perhaps you have finally found a solution to this metadata problem?@ @xizhu1022

[xizhu1022]

Dear author, I like your work very much! Can you provide the data preprocess code (from raw data to the data shown in your paper)? Thank you !

Hi, friend. I have the same the problem, have you had the data preprocess code or an alternative way? Looking for your reply! Thanks!

Hey bro, not yet :( I think you can use Taobao and Beibei dataset, which are commonly used in multi-behavior recommendation (You can refer to the paper GHCF, EHCF).

Thanks for your kind reply. Yelp, the train/test mat files could be simply processed with the open-source datasets and the code context. But you see that the CML model adopted a meta-learning framework that the authors preprocessed a related file which is obscure. In fact, I could not reproduce the experiments without these files or more details. What a pity! @anonymousCML

CML/main.py

Lines 46 to 60 in 00b9a5b

	self.trn_file = args.path + args.dataset + '/trn_'
	self.tst_file = args.path + args.dataset + '/tst_int'
	# self.tst_file = args.path + args.dataset + '/BST_tst_int_59'
	#Tmall: 3,4,5,6,8,59
	#IJCAI_15: 5,6,8,10,13,53
	# self.meta_multi_file = args.path + args.dataset + '/meta_multi_beh_user_index'
	# self.meta_single_file = args.path + args.dataset + '/meta_single_beh_user_index'
	self.meta_multi_single_file = args.path + args.dataset + '/meta_multi_single_beh_user_index_shuffle'
	# /meta_multi_single_beh_user_index_shuffle
	# /new_multi_single
	# self.meta_multi = pickle.load(open(self.meta_multi_file, 'rb'))
	# self.meta_single = pickle.load(open(self.meta_single_file, 'rb'))
	self.meta_multi_single = pickle.load(open(self.meta_multi_single_file, 'rb'))

Hi, friend, perhaps you have finally found a solution to this metadata problem?@ @xizhu1022

No. The authors provided the metadata for IJCAI and Tmall in the SSLRec repo. Even though, how to generate the metadata remains unknown. By the way, I still can not reproduce the experimental results on the IJCAI and Tmall datasets. Actually, there is a significant gap between our reproduced results and the reported results in the paper.