模型效能評估

Machine Learning for Business

Karolis Urbonas

Head of Machine Learning & Science, Amazon

效能評估類型

監督式學習兩大指標:

  1. Accuracy → 分類
  2. Error → 迴歸
Machine Learning for Business

分類效能

  • Accuracy

  • Recall

  • Precision

Machine Learning for Business

流失範例

c1

Machine Learning for Business

流失範例

c2

Machine Learning for Business

流失預測

c3

Machine Learning for Business

錯分項目

c4

Machine Learning for Business

另一個流失預測

c5

Machine Learning for Business

Accuracy(正確率)

acc

Machine Learning for Business

Precision(精確率)

precision

Machine Learning for Business

Recall(召回率)

recall

Machine Learning for Business

迴歸效能

  • Error
Machine Learning for Business

迴歸範例

r1

Machine Learning for Business

用直線預測營收

r2

Machine Learning for Business

迴歸誤差

r3

Machine Learning for Business

測試非線性模型

r4

Machine Learning for Business

降低誤差

r5

Machine Learning for Business

可行模型與 A/B 測試

好模型不一定可執行:

流失預測、購買預測、機器故障預測

測試使用模型是否能改善成效:

對預測將流失的客戶提供誘因(折扣、優惠券、促銷)

對可能購買的客戶寄發提醒信與產品資訊

是否因此降低流失、提高購買率、減少故障?若為 ,納入自動化流程;若為 ,蒐集更多資料、改進模型,再次測試。

Machine Learning for Business

一起來練習吧!

Machine Learning for Business

Preparing Video For Download...