EURO-Online login
- New to EURO? Create an account
- I forgot my username and/or my password.
- Help with cookies
(important for IE8 users)
668. An in-processing and optimization-based method for a fair geographical regression
Invited abstract in session WB-27: Unraveling the Black Box: Advances in Model Explainability, stream Mathematical Optimization for XAI.
Wednesday, 10:30-12:00Room: 047 (building: 208)
Authors (first author is the speaker)
| 1. | Pepa Ramirez Cobo
|
| Universidad de Cádiz |
Abstract
Fairness in Machine Learning is a prominent research area that, however has received little attention in the context of spatial or geographical data. This work is motivated by a real urban planning database that shows lack of equity in terms of access to green areas. On the basis of the geographically weighted regression (GWR), this work presents a regression model for predicting spatial responses in fair way. The parameters of the model, namely, the bandwidth parameter and the regression coefficients are estimated in two stages. First, the bandwidth is obtained from a jack-knife cross-validation approach. Second, the regression coefficients are the solutions of a quadratic optimization problem with linear inequality constraints modeling the unfairness of predicted responses. In order to deal with large datasets, an alternating block coordinate descent algorithm is suggested. The performance of the method is illustrated using an assortment of real datasets.
Keywords
- Machine Learning
- Programming, Quadratic
- OR in Sustainability
Status: accepted
Back to the list of papers