Skip to content

Commit

Permalink
伪标签数据增加说明完善.
Browse files Browse the repository at this point in the history
  • Loading branch information
enjoysport2022 committed Mar 14, 2022
1 parent 540a992 commit 6eccf30
Showing 1 changed file with 3 additions and 1 deletion.
4 changes: 3 additions & 1 deletion autox/autox_competition/feature_engineer/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -90,6 +90,8 @@ lgb_feature_test = featureGbdt.transform(X_test)
获得信息包括:年、月、日、时、一年的第几周、星期、是否工作日、季度、是否月初、是否月末。

# 伪标签数据增强
将测试集中置信度大的样本取出来,用预测结果进行标记,作为增强数据集
使用有标签的数据训练模型, 对无标签的数据进行预测, 将预测置信度大的样本取出来,用预测结果进行标记,作为伪标签增强数据集
### 使用案例
- [PseudoLabeling_AutoX](https://www.kaggle.com/poteman/pseudolabeling-autox)
### 详细说明
- [小数据集怎么上分? 几行代码生成伪标签数据集](https://mp.weixin.qq.com/s/woJCZ-nfK3Bw5cEQfeGJ5A)

0 comments on commit 6eccf30

Please sign in to comment.