Talk:Decision stump

Can somebody just say what it does really simply? I don't know what a "single level decision tree" is, my brain ain't that fancy. Is a decision stump just thresholding on one of the components of a vector or something?Singularitarian (talk) 00:11, 21 April 2010 (UTC)


 * In case of continuous or binary features - it compares value of some feature with a threshold, and outputs one or another value depending on the outcome. I.e. in C code that would be "x[i] < threshold ? value1 : value2", where array x[] contains feature values. -- X7q (talk) 06:45, 21 April 2010 (UTC)

Can someone also explain how decision stump finds the best threshold for each feature? Is there any reference for this classifier other than the Weka source document? — Preceding unsigned comment added by 70.74.162.212 (talk) 04:20, 28 March 2012 (UTC)
 * Well you can simply try all of them and pick the one which minimizes your loss function. There's a finite and rather small number of possible thresholds that lead to different classifications of the training data. Namely try drawing a line between two adjacent points of the training set in each dimension. -- X7q (talk) 23:28, 28 March 2012 (UTC)