threshold_CV: Cross-validation function for tuning threshold in logistic...

In hsansford1/higgsboson:

Description

Function that performs k-fold cross-validation in order to tune the threshold parameter in logistic regression. Fits a weighted logistic regression using randomized training/validation split, then find the the threshold parameter that maximises the approximate median significance (AMS).

Usage

threshold_CV(df, label, weights, theta_0, theta_1, k = 5, n = 200)

Arguments

df	Data-frame to perform cross-validation on.
label	Binary labels of data points in 'df' (0/1).
weights	Weights of data points in 'df'.
theta_0	Lower bound of threshold parameter.
theta_1	Upper bound of threshold parameter.
k	Number of cross-validation sets.
n	Number of values of threshold to check.

Value

'max_theta' is the threshold value that maximizes the AMS and 'max_AMS' is the maximum average AMS found across cross-validation sets.

hsansford1/higgsboson documentation built on Jan. 22, 2022, 4:34 a.m.