But this does only solve the problem for getting a better panorama on a certain classifier, it doesn't oversample the least represented class, so in turn this will only be good for that algorythm. But if I want to make it so the dataset shows a much balanced distribution, I'd go for something like a filter on the data preprocessing tab. The thing is, IDK which one to use and why, could you follow up on this? that'd be great. I'd like to know this because I'm using the experimenter GUI and this wont be of any use there since I'm trying to compare different classifiers to choose the one that would better sort my classes.
@meghnadhalaria27304 жыл бұрын
hello mam excellent explanation, I want to know the range of weight to be set and the difference between cost sensitive learning and cost sensitive classification
@arthurchavescosta12626 жыл бұрын
Nice tool! Thanks. I'm trying to predict models and analyse data from medical exams. I've been trying to find some models (mostly trees) so i could like, draw a profile of each range value of quality index (it's a real value). For that, i divided the index values in value ranges that represent the different classes. So by the end, it was very unbalanced. I'll try this cost matrix to see if it adjusts better. When using linear regression to show how much the quality index value depends on each of the attributes, i came across a litlle question: If the data was normalized, would it significant for the results?...
@datascientist29584 жыл бұрын
Can we implement XGB (Extreme Gradient Boosting) in Weka?