Does anyone have an idea of what the predicted value of regression tree would be when the feature takes a value at the point where a split happens. Based on the example explained in module 5, the regression tree could yield a range of predicted values when the feature value occurs at a split.
By inspecting the illustration of the tree with max_depth=1 in the Decision tree for regression notebook, you can notice that the intervals are defined by "Flipper Length (mm)" <= 206.5 vs "Flipper Length (mm)" > 206.5.
So a sample that lies exactly on the split is assigned to the left. This choice is somewhat arbitrary.