Reputation: 33
How to increase the weight of observation in boosting regression
How to increase the weight of observation in boosting regression, like for boosting classification, you could add more weights to the observations who predict wrong but for boosting regression, how could you know which observation is wrong and add weights towards them? Do we have an error range for boosting regression or so?
Upvotes: 0
Views: 214
Answers (1)
Reputation: 1
This is what https://scikit-learn.org/stable/modules/generated/sklearn.ensemble.AdaBoostRegressor.html do. Refer: Improving Regressors using Boosting Techniques.
Actually, Every step of boosting you have a loss, for example loss=(y-f(x))**2, then you can use loss to represent weight. Weight can be used to two ways. One is using weight to resample dataset. The other is using weight directly in loss, in optimization
Upvotes: 0
Related Questions
- Improving boosting model ,reducing Root mean square error
- How to assign feature weights in XGBClassifier?
- How to use adaboost with different base estimator in scikit-learn?
- How can I tweak xgboost to assign more weight to a variable?
- Upweight Or adding weight to the downsampled examples
- Gradient Boosting classifier issue
- setting booster in a XGBClassifier
- Use of sample_weight in gradient boosting classifier
- How to add weight vector to xgboost when calling via pandas_ml
- adjust weights for predicted classes in xgboost in loss function