Micro-Macro Precision,Recall and F-Score
I was always mix-up precision and recall and this messes up things more times the you can imagine !!.So thought of writing a short post to summarize an easy interpretable definition of the following terms:
Precision
Recall
F-Score
To understand the above we require the knowledge of True/False Positives and Negatives which can be easily be remembering the following confusion matrix
Now, that we get what True/False Positives & Negatives are we can formally define
Precision as fraction of true positive among all the positive’s recalled.
Recall as fraction of true positives among all the correct events.
F-Score (Balanced ) as the harmonic mean of the Precision and Recall
Micro Averaged metrics given two different set of data :
Macro Averaged metrics with two datasets :
USAGE:
Macro-averaged metrics are used when we want to evaluate systems performance across on different datasets.
Micro-averaged metrics should be used when the size of datasets are variable.
This is my first post ,hope you find it useful !! 😃