Assembled by Sohyeon Yim (2020-06-28)
- 모델이 1이라고 분류해낸 그룹 A가 있을 때, 모델이 얼마나 믿을 만한 정도로 A를 만들어 냈는지 평가
- 모형의 실용성과 관련된 척도
- general formula for positive real β, where recall is considered β times as important as precision
- 실무에서는 F1-measure보다 F-beta measure를 더 많이 쓴다. precision이나 recall에 가중치를 주고 싶을 때, 이 식을 사용하면 된다. precision에 가중치 3을 주고 싶으면 β에 1/3을 대입하면 된다. 한편 recall에 가중치 3을 주고 싶으면 β에 3을 대입하면 된다.
- more weight on precision, less weight on recall
- less weight on precision, more weight on recall
A Gentle Introduction to the Fbeta-Measure for Machine Learning