In this article, we propose a Weighted Oblique Logistic Tree for Regression (WOLTReg), a novel method to build regression oblique trees using a weighted logistic regression classifier to determine the optimal oblique splitting hyperplanes; this novel approach at each node integrates a weighting mechanism to reduce impurity and a feature selection process that avoids data overfitting. We also evaluate WOLTReg on multiple benchmark datasets and find out that WOLTReg works better when only a small subset of predictors is used to create the separating hyperplane. Furthermore, we compared its performance in a test set across all datasets with other oblique tree methods and a standard decision tree. The results indicate that WOLTReg has better performance with respect to the other methods in terms of root mean square error. In conclusion, this study highlights the potential of WOLTReg and suggests future applications in various domains.