Today I Learned - 211104

Interpreting ML Model

model-agnostic이란?

모델에 의존하지 않고, input 과 output을 바탕으로 설명하는것.

PDP (Partial Dependence Plot)

  • easy to understand, but assumes no correlation between features
  • describes between model inputs and outputs
  • 피팅된(학습된) 모델에 적용하여, 각 feature들이 결과에 어떤 영향을 주었는지 조사하는 방법 중 하나이다.
  • Feature importance는 variable 중 예측에 중요하게 작용한 것이 무엇인지 보여준다면, Partial dependence는 각 variable이 예측에 어떻게 작용했는지 보여준다.

SHAP (SHapley Additive exPlanations)

  • Breaks down a prediction to show the impact of each feature.
  • 어떠한 feature가 갖는 baseline value에 비해 해당 모델에서 어떤 value를 갖는지 보여줌.

We could ask:

  • How much was a prediction driven by the fact that the team scored 3 goals?

But it’s easier to give a concrete, numeric answer if we restate this as:

  • How much was a prediction driven by the fact that the team scored 3 goals, instead of some baseline number of goals.

img


© 2023. All rights reserved.