The experimental branch looked to solve the overfitting problem, described in issue #21. Fix is not nominating algorithms for "best algorithm" if their balanced accuracy less than or equal to 50%, |sensitivity-specificity| greater than 0.85, sensitivity equaling 0 or 1, or specificity equaling 0 or 1. If none meet these requirements, then best algorithm is nominated as previously done before (best based on the metric the user chooses to maximize) (thanks @jfcarter2358 for the help!)
Branch also includes a new way to plot the probability plots that will break up each case/class type into separate lines, this was courtesy of @mikeDTI (thanks, Mike!)
Both these features will be in the next genoml2 package release 🥳
The experimental branch looked to solve the overfitting problem, described in issue #21. Fix is not nominating algorithms for "best algorithm" if their balanced accuracy less than or equal to 50%, |sensitivity-specificity| greater than 0.85, sensitivity equaling 0 or 1, or specificity equaling 0 or 1. If none meet these requirements, then best algorithm is nominated as previously done before (best based on the metric the user chooses to maximize) (thanks @jfcarter2358 for the help!)
Branch also includes a new way to plot the probability plots that will break up each case/class type into separate lines, this was courtesy of @mikeDTI (thanks, Mike!)
Both these features will be in the next genoml2 package release 🥳