Data Leakage in Machine Learning: Why a 95% AUC Can Collapse to 68% in Production
中文摘要
本指南探讨了时间、目标、重复数据泄漏及数据漂移如何导致机器学习模型在生产环境中的性能大幅下降。
English Summary
This guide explains how data leakage and drift cause high-performing machine learning models to collapse in real-world production.
原文节选
A practical guide to temporal leakage, target leakage, duplicate leakage, data drift, label shift, and point-in-time feature engineering. Continue reading on Medium »