I need to predict the next target event – buying a car in two price categories (for example, high class and middle class). The number of targets for training is approximately the same (let there be 10 thousand purchases in each price category). I see two approaches to training machine learning models here. The first is a multiclass model. The second is two separate models to predict each segment. Which approach do you think should be taken and why? I want to take the multi-class model as a basis, what advantages and disadvantages can you list?