Question about the evaluation metrics in traj_loc_pred task

[SunderlandAJ-1130]

Dear all, thanks for your useful codes.

I have a small question about the evaluation metric about Precision@K in your codes. I am not sure why you only use "self.intermediate_result['total'] * self.topk" to calculate Precision@K and remove "self.topk" in the remaining metrics (e.g., Recall@K).

Thanks!

https://github.com/LibCity/Bigscity-LibCity/blob/f8e5e49a162cb4c5660f655b43d9c0a919a21193/libcity/evaluator/traj_loc_pred_evaluator.py#L62-L71

[WenMellors]

Hello, these metrics are calculated as described in the document https://bigscity-libcity-docs.readthedocs.io/en/latest/user_guide/evaluator/traj_loc_pred.html. You can look up "Recall and Precision at k for Recommender Systems" to understand how these metrics are computed. In essence, you treat the actual next location visited by the user as the item they purchased, and the model's top k predicted locations as the top k recommended items.

[SunderlandAJ-1130]

Thanks for your generous explanation. We will cite your paper correctly in our current study. Thanks!