Visualizing Machine Learning Thresholds to Make Better Business Decisions

[sl8r]

http://blog.insightdatalabs.com/visualizing-classifier-thresholds/

[baq]

this is one of many cases in real life where a naked average is suboptimal. one should always analyze variance, median, some centiles (pick a couple you like), min/max and if data is amenable to that, scatter plots - if your data is e.g. bimodal, aggregates usually won't tell you.