检测数据与模型漂移
发现现实环境何时发生变化
检测数据与模型漂移 是 CoddyKit 上的免费 Deep Learning Academy 课时。 这是第 3 节课，共 4 节。 你可以在下方免费阅读本课时的完整内容 — 然后在浏览器中使用内置代码编辑器和全天候 AI 导师进行实践。 这是 Deep Learning Academy 学习路径的一部分，你的进度在网页和 CoddyKit 应用中同步。 Deep Learning Academy 课程共包含 4 节课。
本课时的部分内容尚未翻译，以英文显示。
The World Keeps Moving
Your model learned yesterday's patterns, but real data keeps changing. When it shifts, predictions quietly get worse. We call this problem drift. 🌊
Two Flavors of Drift
There are two kinds to watch. Data drift means the inputs change shape, while concept drift means the relationship between inputs and labels changes.
A Concrete Example
A spam filter trained on old emails sees new tricks it never met. The inputs drifted, so its once-sharp accuracy slowly decays in production.
Watch the Input Distribution
The first signal of drift is in the features. Compare the distribution of recent inputs to your training data and look for a shift in the shape.
ref_mean = X_train.mean(axis=0)
live_mean = X_live.mean(axis=0)Measure the Gap
To quantify how far two distributions diverge, use a statistic like KS, the Kolmogorov-Smirnov test, which returns a number plus a p-value.
from scipy.stats import ks_2samp
stat, p = ks_2samp(X_train[:,0], X_live[:,0])Set a Threshold
A small p-value means the live data no longer matches training. Pick a threshold and trigger an alert whenever the gap crosses it.
if p < 0.05:
print("drift detected")Watch Predictions Too
Even without labels, the spread of your model's outputs is a clue. A sudden shift in predicted probabilities often signals incoming drift.
The Best Signal Is Truth
When real labels eventually arrive, compare them to past predictions. Falling live accuracy is the clearest, most direct proof that drift is hurting you.
Tools That Watch for You
You need not build everything by hand. Libraries like Evidently compute drift reports across all features and flag the columns that moved.
from evidently.report import ReportMonitor Continuously
Drift is not a one-time check. Run these comparisons on a schedule so your monitoring catches slow shifts long before users complain.
Detection Is Half the Battle
Spotting drift early lets you act before damage spreads. The natural next move is to retrain on fresh data and restore the model's edge.
Quick Check
What does concept drift specifically describe?
Recap
You learned to spot drift: compare live inputs to training, measure the gap with tests, watch predictions, and monitor on a schedule. 🎉
常见问题解答
「检测数据与模型漂移」课时是免费的吗？
是的 — 「检测数据与模型漂移」的完整文本可在网页上免费阅读。要进行交互式练习（内置代码编辑器和全天候 AI 导师）并解锁 Deep Learning Academy 课程的其余内容，请升级到 CoddyKit PRO。 Deep Learning Academy 课程共包含 4 节课。
「检测数据与模型漂移」这节课中我会学到什么？
发现现实环境何时发生变化 你通过在浏览器中直接运行的动手代码来练习 Deep Learning Academy，全天候 AI 导师会在你学习这节课的过程中回答你的问题。
学习 Deep Learning Academy 需要有经验吗？
无需任何先前经验。CoddyKit 上的 Deep Learning Academy 课程适合初学者到高级学习者，你可以从这里开始或从头开始，按照自己的节奏学习。 这是第 3 节课，共 4 节。
「检测数据与模型漂移」课时需要多长时间？
大多数 CoddyKit 课程大约需要 5–10 分钟。每节课都很精短且互动，所以你能稳步进步，并在网页和应用中从离开的地方继续。
我能在这节 Deep Learning Academy 课中编写并运行代码吗？
能。每节 Deep Learning Academy 课都包含内置代码编辑器，你可以在浏览器中直接编写并运行真实代码，并获得即时 AI 反馈 — 无需本地设置。