B1016
Title: Feature selection in mixture of logistic regression models using the modified elastic-net penalty
Authors: Salomi Millard - University of Pretoria (South Africa) [presenting]
Sollie Millard - University of Pretoria (South Africa)
Mohammad Arashi - Ferdowsi University of Mashhad (Iran)
Frans Kanfer - University of Pretoria (South Africa)
Gaonyalelwe Maribe - University of Pretoria (South Africa)
Abstract: Datasets with a relatively large number of highly correlated features are often found in applications of finite mixture regression models. Furthermore, the contribution of each feature towards the response variable differs in the respective components of the mixture model. This creates a complex feature selection problem. Penalised regression methods are frequently used to perform feature selection whilst addressing the issues that arise due to multicollinearity. We consider the use of the novel modified elastic-net (MEnet) penalty for statistical analysis and feature selection in a finite mixture of logistic regressions setting. A simulation study is performed to demonstrate the properties pertaining to feature selection and classification accuracy of this approach.
