多目标的回归问题(MTR, Multiple Target Regression)
Why use multi-output models instead of a combination of single-output models like regression?
- Models using single outputs take longer to train and are computationally expensive
- They optimize for the single target rather than all the targets together
- They do not use relationships between the target variables (Y)
- MTR regression models are simpler than a stack of Single Target models
- MTR trees provide human-readable predictive models and the models are easily interpreted.
参考资料
https://towardsdatascience.com/regression-models-with-multiple-target-variables-8baa75aacd