Improving SVM-linear predictions using CART for example selection

• Author(s): João Pedro Carvalho Leal Mendes Moreira ; Alípio Mário Guedes Jorge ; Carlos Manuel Milheiro de Oliveira Pinto Soares ; Jorge Rui Guimarães Freire de Sousa
• Date: 2006
• Persistent ID: http://hdl.handle.net/10216/50012
• Origin: Repositório Aberto da Universidade do Porto
• Subject(s): Ciências tecnológicas; Tecnologia; Tecnologia dos transportes
• Document Type: Conference Object
• Language: English

Description

This paper describes the study on example selection in re- gression problems using ¹-SVM (Support Vector Machine) linear as pre- diction algorithm. The motivation case is a study done on real data for a problem of bus trip time prediction. In this study we use three differ- ent training sets: all the examples, examples from past days similar to the day where prediction is needed, and examples selected by a CART regression tree. Then, we verify if the CART based example selection ap- proach is appropriate on di®erent regression data sets. The experimental results obtained are promising.