CODEWITHSUNDEEP
pythonmachine-learningscikit-learnconfusion-matrix
zeal
zeal

Reputation: 485

Confusion with SKlearn Precision-Recall Curve computation

Following is the snippet from sci-kit pr-curve computation. 

>>> import numpy as np
>>> from sklearn.metrics import precision_recall_curve
>>> y_true = np.array([0, 0, 1, 1])
>>> y_scores = np.array([0.1, 0.4, 0.35, 0.8])
>>> precision, recall, thresholds = precision_recall_curve(
...     y_true, y_scores)
>>> precision  
array([ 0.66...,  0.5       ,  1.        ,  1.        ])
>>> recall
array([ 1. ,  0.5,  0.5,  0. ])
>>> thresholds
array([ 0.35,  0.4 ,  0.8 ])

Doubts:

Why are thresholds only 3 while precision and recall given are 4. As one can clearly see the threshold of 0.1 is left out. And the computation starts from threshold 0.35 and more.

Upvotes: 1

Views: 1253

Answers (1)

Bert Kellerman
Bert Kellerman

Reputation: 1629

The thresholds only go low enough to attain 100% recall. The idea being that you generally wouldn't set a lower threshold as it would introduce unnecessary false positives.

https://github.com/scikit-learn/scikit-learn/blob/a24c8b46/sklearn/metrics/ranking.py

  

   # stop when full recall attained
   # and reverse the outputs so recall is decreasing
    last_ind = tps.searchsorted(tps[-1])            
    sl = slice(last_ind, None, -1)
    return np.r_[precision[sl], 1], np.r_[recall[sl], 0], thresholds[sl]

Upvotes: 2

Related Questions