Stability selection for component-wise gradient boosting in multiple dimensions
arXiv:1611.10171 · doi:10.1007/s11222-017-9754-6
Abstract
We present a new algorithm for boosting generalized additive models for location, scale and shape (GAMLSS) that allows to incorporate stability selection, an increasingly popular way to obtain stable sets of covariates while controlling the per-family error rate (PFER). The model is fitted repeatedly to subsampled data and variables with high selection frequencies are extracted. To apply stability selection to boosted GAMLSS, we develop a new "noncyclical" fitting algorithm that incorporates an additional selection step of the best-fitting distribution parameter in each iteration. This new algorithms has the additional advantage that optimizing the tuning parameters of boosting is reduced from a multi-dimensional to a one-dimensional problem with vastly decreased complexity. The performance of the novel algorithm is evaluated in an extensive simulation study. We apply this new algorithm to a study to estimate abundance of common eider in Massachusetts, USA, featuring excess zeros, overdispersion, non-linearity and spatio-temporal structures. Eider abundance is estimated via boosted GAMLSS, allowing both mean and overdispersion to be regressed on covariates. Stability selection is used to obtain a sparse set of stable predictors.
16 pages
References in corpus (5)
- Boosting Algorithms: Regularization, Prediction and Model Fitting
- The Evolution of Boosting Algorithms - From Machine Learning to Statistical Modelling
- Variable selection with error control: Another look at Stability Selection
- Extending Statistical Boosting - An Overview of Recent Methodological Developments
- Spatiotemporal modelling of sea duck abundance: implications for marine spatial planning
Cited by in corpus (6)
- A review of predictive uncertainty estimation with machine learning
- XGBoostLSS -- An extension of XGBoost to probabilistic forecasting
- Boosting Functional Regression Models with FDboost
- CatBoostLSS -- An extension of CatBoost to probabilistic forecasting
- Extension of the Gradient Boosting Algorithm for Joint Modeling of Longitudinal and Time-to-Event data
- Random boosting and random^2 forests -- A random tree depth injection approach