id	author	title	date	pages	extension	mime	words	sentences	flesch	summary	cache	txt
work_kdo3jxr3xvh7rbksrlfmoi3qxe	Luís Torgo	Resampling strategies for regression	2014	20	.pdf	application/pdf	7957	694	70	Several real world prediction problems involve forecasting rare values of a target variable. Namely, we propose to address such tasks by re-sampling approaches that change the distribution of the given data set to decrease These modifications allow the use of these strategies on regression tasks where the goal is to forecast rare extreme values of the approaches for the task of predicting rare extreme values of a numeric target variable. is that the goal is the predictive accuracy on a particular subset of the domain of the target variable Y the rare and extreme values. Namely, we propose the strategy of under-sampling observations whose target value has a relevance less than a user-defined parameter. The original Smote algorithm uses an over-sampling strategy that consists on generating "synthetic" cases with a rare target value. sampling approaches at predicting rare extreme values of a continuous target	./cache/work_kdo3jxr3xvh7rbksrlfmoi3qxe.pdf	./txt/work_kdo3jxr3xvh7rbksrlfmoi3qxe.txt