Thanks for your wonderful work. Well-organized project!
My experimental results show that by replacing the original E_Day with the item's categories, both the Recall and NDCG metrics have significantly improved. My goal is to reduce the use of time-related information and increase the use of item category information, allowing the model to learn the user's item category trends. My approach to handling categories is as follows, using ML-1M as an example:
Animation|Children's|Comedy 0
Adventure|Children's|Fantasy 1
Comedy|Romance 2
...
Of course, you can try other handling methods.
My experimental results are as follows:
Thanks for your wonderful work. Well-organized project!
My experimental results show that by replacing the original E_Day with the item's categories, both the Recall and NDCG metrics have significantly improved. My goal is to reduce the use of time-related information and increase the use of item category information, allowing the model to learn the user's item category trends. My approach to handling categories is as follows, using ML-1M as an example:
Animation|Children's|Comedy 0
Adventure|Children's|Fantasy 1
Comedy|Romance 2
...
Of course, you can try other handling methods.
My experimental results are as follows:
