Apr, 2023
机器学习模型训练中对表现较好少数群体过度采样略微减少不良影响但也降低了模型准确性
Oversampling Higher-Performing Minorities During Machine Learning Model Training Reduces Adverse Impact Slightly but Also Reduces Model Accuracy
Louis Hickman, Jason Kuruzovich, Vincent Ng, Kofi Arhin, Danielle Wilson
TL;DR该研究使用机器学习模型对人事评估进行建模,探讨训练数据中负面影响比率对模型预测结果的影响,发现训练数据中负面影响比率与模型异常影响呈线性关系,但是从训练数据中去除负面影响只能略微减少异常影响,同时会对模型准确性产生负面影响。