Open Chuan1997 opened 1 month ago
您好,我想请问如果是实现overall user/item recommendation,pipeline是怎么样的?还有我有自己的数据集,已经划分好了,如何load呢?
Hi, thanks for the question. The model testing has already included cold, warm, and overall user/item recommendations. The evaluation for overall user/item recommendations followed ALDI's way (https://dl.acm.org/doi/pdf/10.1145/3539618.3591732), which the recommendation will include both warm and cold sets. You can see ALDI's paper and convert.py in ./data folder for more details (we will add more code comments in the next update).
For your own datasets, if you have processed in your way, you should convert the format into the .csv file the same as ColdRec (like the processed data in https://drive.google.com/drive/folders/13JJ25vf5dpFzxe1ITQYONrEIUAsB2ZU8). Then, you can load the same as ColdRec's data loading module. If you have unprocessed raw datasets, you can directly process them with our provided files in ./data folder for unified splitting and usage.
Hope this could help you.
Thanks for the timely response. Now the code perfectly runs on my datasets, and I have another question, to get the perfect performance on the dataset, the first step is to run param_search.py finding the best parameters, and then run the official test?
And also, if I try to run the warm recommender, there will be a mistake like:
File "/home/coldrecown/config/model_param.py", line 55, in model_specific_param
raise Exception("The model name not found.")
Exception: The model name not found
I want to ask if you would like to communicate via WeChat for convenience, my account is ywc1533293533.
您好,我想请问如果是实现overall user/item recommendation,pipeline是怎么样的?还有我有自己的数据集,已经划分好了,如何load呢?