1. Data Imbalance

  • resampling;cost-sensitive:迭代优化cost matrix,类间损失,根据每个epoch的训练错误调整loss权重(re-weighting),和五元组的combine,考虑sub-cluster的weight,iterative weighting matrix
  • under-sampling with bagging;NearMiss(数据清洗);over-sampling:SMOTE;

1.1. Reference

results matching ""

    No results matching ""