EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
1481. Selecting variables in additive models as linear or non-linear terms
Invited abstract in session TD-27: Feature attribution and selection for XAI, stream Mathematical Optimization for XAI.
Tuesday, 14:30-16:00Room: 047 (building: 208)
Authors (first author is the speaker)
| 1. | Vanesa Guerrero
|
| Department of Statistics, Universidad Carlos III de Madrid | |
| 2. | Manuel Navarro García
|
| Departamento de Estadística, Universidad Carlos III de Madrid | |
| 3. | Maria Durban
|
| Universidad Carlos III de Marid | |
| 4. | Arturo del Cerro
|
| Universidad Carlos III de Madrid |
Abstract
In an era when the decision-making process is often based on the analysis of
complex and evolving data, it is crucial to have systems which allow us to incorporate human knowledge and provide valuable support to the decision maker. In this work, statistical modelling and mathematical optimization paradigms merge to address the problem of feature selection in additive models while deciding if the variables are linear or nonlinearly related with the outcome. We assume that the smooth functions involved in the additive model are defined through a reduced-rank basis (B−splines) and fitted via a penalized splines approach (P−splines). A mixed-integer quadratic optimization model is proposed to address the variable selection model and a matheuristic approach, based on the Large Neighborhood Search paradigm, is developed to solve larger instances. The proposed methodology is tested in both simulated and real datasets and it is shown to be competitive against other approaches in the literature.
Keywords
- Analytics and Data Science
- Programming, Mixed-Integer
- Metaheuristics
Status: accepted
Back to the list of papers