Modelo para suporte à descoberta de conhecimento em base dedados (KDD): aplicação em estratégias de venda no mercado de medicina diagnóstica

Abstract

The great deal of data stored in databases of healthcare organizations may mask valuable and useful possibilities in decision making. The health sector offers numerous possibilities for applications of these techniques due to the complexity of the processes and the large storage volume of your data in use by information systems. In a specific domain, a challenge of the information retrieval process is to create the semantic relation between the terms of a specialized vocabulary. A wellknownalternative to identify hidden standards is the use of data-mining techniques. In order to obtain more efficiency in data-mining, ontologies have been used to Such study aims at describing a proposal for the use of biomedical ontologies diseases and laboratory tests on the process with Knowledge Discovery in Database (KDD) to make more effective information retrieval on the prescribing behavior of laboratorytests which is, in this case, related to viral hepatitis. It is associated with such knowledge organizational technique and its data mining which contributes to the information retrieval and, consequently, to the knowledge gained. The model developed shows ontology additional testing of viral hepatitis to generalize its attributes, in the pre-processing phase, and in the pos-processing phase and classify the association rules obtained considering the semantic similarity between the antecedent and consequent, assess additional testing considering whatsimilarly relates to their disorders. The domain ontologies are used to introduce its theoretical terms and introduce experts knowledge which allows its patterns control and assessment, considering the most interesting items. After model validation, a database was created with the January to March of 2015 requests, which contained some additional testing directly related to the diagnosis of viral hepatitis. The records to be mined were widespread by an application developed in Java within Jena framework and subsequently the association rules obtained wereclassified using the Tversky similarity calculation model. The results showed that with the generalization of attributes, the consolidation of related additional tests, and subsequently the classification of rules, evaluating the features related to the disease, its symptoms and where they prevailingly attack using ontology is possible to reduce the number of generated patterns and therefore recover more relevant information in decision making. The analysis of the results allows managers to target more effectively their marketing sales cycles and sales approach.