Por favor, use este identificador para citar o enlazar este ítem:
http://cimat.repositorioinstitucional.mx/jspui/handle/1008/1117| LOCAL SAMPLINGSVM AND MONORITY OVERSAMPLING BOOSTING : SAMPLING-BASED APPROACHES FOR STATISTICAL | |
| Roberto Bárcenas | |
| Acceso Abierto | |
| Atribución-NoComercial | |
| PROBABILIDAD Y ESTADÍSTICA | |
| This thesis addresses problems of statistical classification, with emphasis on data obtained on a large-scale and with an unbalanced distribution in the classes. The main objective is to introduce alternative classification algorithms based on well-known approaches, such as Support Vector Machines and Boosting, to solve specific issues that exist in different important applications. We consider the local sampling method to find SVM solutions and the use of informed oversampling in a boosting scheme to face the imbalanced classification scenario. One of the specific goals is to reduce computation time by pre-processing the data, allowing us to identify subsets of interest that have a reduced size and sufficient information, where it is also possible to replicate the task of applying classification techniques. The extended aim of this research is to develop necessary theoretical foundations in order to assess the relevance of these solutions, which means establishing a probabilistic manner the properties of the algorithms presented. We demonstrate that under some conditions, our proposals are relevant by using objective criteria and evaluating them through performance measurements and, also, we obtained benefit in terms of computational costs when it is compared to the standard techniques available. | |
| 17-01-2020 | |
| Trabajo de grado, doctorado | |
| OTRAS | |
| Versión aceptada | |
| acceptedVersion - Versión aceptada | |
| Aparece en las colecciones: | Tesis del CIMAT |
Cargar archivos:
| Fichero | Descripción | Tamaño | Formato | |
|---|---|---|---|---|
| TE 820.pdf | 4.79 MB | Adobe PDF | Visualizar/Abrir |