logistic regression - R - Stepwise random results -
if set seed before regression (logistic)
lr = glm(target ~ 1, family=binomial, data = x) set.seed(12345) lr = stepaic(lr, scope = f, family=binomial, data = x, k = 3) summary(lr)
i different results depending on seed different performance. 1 random part of algorithm? because in neural net result depends on first weights if have local minima, in logistic regression random part?
(since conceptual question didn't add reproducible example.)
Comments
Post a Comment