Recall vs Precision vs F1

ref: youtube.com/watch?v=2osIZ-dSPGE&list=PL..

Accuracy - Doesnt matter positive or negative prediction (true positive / all data samples)

Precision - true positive / (false positive + true positive) [aka all positive]

for precision, think about predictions as your base (datapoints that are predicted true)

Recall - true positive / (false negative + true positive) [aka truth samples]

for recall, think about truth as your base (datapoints that are real true)

Confusion matrix - supply truth & prediction, then will plot true positive, false positive, true negative, false negative

F1 gives you an overall performance of your model (combination of precision & recall)