試験
通常
試験時間: 00:00問題時間: 00:00
問題一覧
問題 64 / 65
When you suspect data leakage because only some departments show large performance differences between training and evaluation data, which validation design should you check first?
①Create an ensemble of models to improve stability.
②Deploy the model to production immediately in accordance with the policy.
③Modify the evaluation metric and re-evaluate.
④Delete inference logs to clean the dataset.
⑤Validate using cross-validation and external holdouts that take department and time into account.