Um método baseado em hipóteses estatísticas para a descoberta de itemsets com distribuição assimétrica em processos de mineração de dados.

AUTOR(ES)
DATA DE PUBLICAÇÃO

2005

RESUMO

Nowadays, the ability to make good decisions at the right moment is a very desirable characteristic. To assist in the process of decision making, various techniques have been created. These techniques, however, need useful and reliable information to be eective. Data mining is an essential step in knowledge discovery in databases, being thus very important to decision making processes. This work presents a method based on statistical hypotheses for discovering itemsets with asymmetrical distribution. Itemsets with asymmetrical distribution can generate association rules that do not represent all database, but only a fraction of it. This method extracts class-dependent rules, which are association rules generated from itemsets whose distribution between the semantically distinct sets of the database is asymmetrical. These rules make explicit the domain which the association rule really represents. The presented method is applicable to centralized and distributed databases, and can also be used to increase the reliability of association rules found in dynamic databases. The method is capable of discovering additional information about association rules that cannot be discovered by traditional data mining algorithms. This additional information is of great value to decision making processes, because it indicates if an association rule can be considered globally valid and if it does not possess unexpected responses.

ASSUNTO(S)

distribuição (probabilidades) metodologia e tecnicas da computacao testes de hipóteses estatísticas data mining (mineração de dados) banco de dados - gerência

Documentos Relacionados