Can I tell a machine learning model that the dependent variable is normally distributed?

Question

I am trying to set up a machine learning model predicting a continuous variable y on the basis of a feature vector (x1, x2, ..., xn). I know from elsewhere that y follows a normal distribution. Can I somehow specify this to the model and enhance its predictions this way? Is there a specific model that allows me to do this?

I have used linear models, k-nearest neighbour models and random forest models (in python). All of them give some predictions but I was wondering whether they can be outperformed by some model that would know the distribution of the predicted variable.

Answer 1

做到这一点的一种方法是创建一个自定义目标函数，该函数对未按正态分布的预测进行惩罚。