WIREs Data Mining Knowl Discov
Predictive data mining in clinical medicine: a focus on selected methods and applications
Can't access this content? Tell your librarian.
Biomedical informatics is a discipline with several domains of applications. It deals with diverse data sources including molecular and cellular processes, tissues and organs, and individual patients and populations (adapted from Ref 1). Data mining can be successfully applied in all areas to support decision‐making activities.
Learning clinical predictive models requires a careful evaluation process. Different data sources need to be properly integrated and preprocessing and feature selection may turn out to be the most important parts of data analysis. Model evaluation requires an independent data set to assess the prediction performance. Finally, the model should be deployed carefully taking into account the clinical context.
Clinical decision‐making is largely based on results coming from clinical trials. The analysis of clinical databases performed with data mining approaches provides a way to generate hypotheses for further trials and suggest changes in day‐by‐day practice based on the accumulated experience.