Aprendizagem de máquina baseada na combinação de classificadores em bases de dados de área de saúde

AUTOR(ES)
DATA DE PUBLICAÇÃO

2007

RESUMO

Nowadays most decision problems do not have as a challenge the numeric treatment, but the transformation of data and information into knowledge, specially, when data bases are related with health. These health data bases, in general, have many attributes, few instances and many missing values, which, regarding machine learning, leads to redundant and irrelevant data. The main purpose of this work is the experimentation of a simple machine learning method (J48) and two more sophisticated methods (BAGGING e BOOSTING) to health data bases, in order to verify the efficiency of these methods and suggest efficient solutions for knowledge discovery. The methods efficiency was evaluated by the generation of learning curves to each method over the same set of health data bases. Another part of this work is the analysis of the impact of the number of combined classifiers in the application of methods BAGGING and BOOSTING. Another important point of this work was the application of experiments in data bases in their original format and also the same bases submitted to an attribute selection technique. As for this work contribution, an analysis of the experiments is presented leading to the recommendation of one of the machine learning methods. Evidently, such recommendation is valid only for data bases with the same characteristics as those used in this work.

ASSUNTO(S)

aprendizado do computador engenharia medica exploração de dados medicina - processamento de dados machine learning medicine biomedical engineering engenharia biomédica - dissertações data mining

ACESSO AO ARTIGO

Documentos Relacionados