Full text
| |
| Author(s): |
Jaqueline Brigladori Pugliesi
Total Authors: 1
|
| Document type: | Doctoral Thesis |
| Press: | São Carlos. |
| Institution: | Universidade de São Paulo (USP). Instituto de Ciências Matemáticas e de Computação (ICMC/SB) |
| Defense date: | 2004-05-07 |
| Examining board members: |
Solange Oliveira Rezende;
Alexandre Gonçalves Evsukoff;
Maria Carolina Monard;
Jaime Simão Sichman;
Flavio Soares Correa da Silva
|
| Advisor: | Solange Oliveira Rezende |
| Abstract | |
Data Mining process begins with the understanding of the application domain, considering aspects as application objectives and data sources. Then, the data pre-processing and pattern extraction is realized. After the pattern extraction stage, one proceeds with the post-processing, in which the knowledge is evaluated as regards its quality and/or usefulness in order to use this knowledge to support a decision making process. Recently, much attention has been given to regression problems. However, regression in predictive Data Mining is a little explored subject in the knowledge discovery from database process, what makes the study of exploration methods very relevant. Some work in areas related to the Knowledge Discovery of Data and Text Bases process have been accomplished at LABIC (Laboratório de Inteligência Computacional) which motivated the construction of a computational environment for knowledge extraction called DlSCOVER. The WfnPFL environment was proposed and developed to aim the symbolic regression model construction and the regression problems post-processing, This environment makes possible the evaluation of regression rules, providing strategies for contingency table calculation and the subsequent utilization of ali measures derived from this table for regression rules evaluation. Moreover, the system also provides the combination of homogeneous and heterogeneous regressors to improve the regressor precision and the integration and pruning of regression rules obtained from different samples or algorithms. These functionalities of Ú&&PE increase the DlSCOVER potentiality in relation to regression treatmet. (AU) | |