Precision / Recall 及 F1-score

 

True positive(tp):　algorithm predicts 1 and it actually is 1

True negtive(tn):　algorithm predicts 0 and it actually is 0

False positive(fp):  algorithm predicts 1 and it actually is 0

False negative(fn): algorithm predicts 0 and it actually is 1

 

Precision: 在所有预测为1结果里面有多少是真实为1

$precision = \frac{True\ positives}{\#predicted\ positives} = \frac{tp}{tp + fp}$

 

Recall: 所有真实为1的有多少能够被成功预测为1

$recall = \frac{True\ positives}{\#actual\ positives} = \frac{tp}{tp + fn}$

 

F1-score: 偏斜集衡量指标

$F_{1}\textrm{-}score = \frac{2PR}{P+R}$

 

转载于:https://www.cnblogs.com/freyr/p/4607997.html