News for hackers. Stuff that matters.
Re:Counting solar panels in the U.S. with machine learning a... by desdiv on Friday February 01, 2019 @04:06AM (#19052699)
Going by this graph[0], what they meant was: we had a precision of 93% and a recall of 90%.

[0] http://web.stanford.edu/group/deepsolar/assets2/img/roc.jpg

Re:Counting solar panels in the U.S. with machine learning a... by floatrock on Friday February 01, 2019 @11:12AM (#19055325)
Wikipedia has a good illustration showing the relation between precision & recall vs. false positives & false negatives:

https://en.wikipedia.org/wiki/Precision_and_recall

Precision is what percent of the identified-positives are actually positives: true-positives / (true-positives + false-positives).

Recall is what percent of the actual-positives were identified as positive: true-positives / (true-positives + false-negatives).

These are useful summary stats over the true-false/positive-negative matrix because, to quote the wikipedia: "In simple terms, high precision means that an algorithm returned substantially more relevant results than irrelevant ones, while high recall means that an algorithm returned most of the relevant results."