Knowledge guided data mining

The information contained in many databases can be extremely valuable for a variety of complex decisions. However, often that information is hidden or buried under the volume and organisation of the data in the database. In order to find little nuggets of gold of useful information, it is necessary to conduct a process of data mining. This process uses induction techniques guided by knowledge in order to identify important trends, characteristics or decision factors within a database. The purpose of induction in this context is to intelligently split the data sample into two or more homogeneous data sets (or clusters) based on features present within the data. Since we are basically searching for a Boolean condition (the outcome is either met or it is not met), there will be two types of cluster: those which predominantly contain the outcome, and those which predominantly do not contain the outcome. Induction would then provide us with a prioritised set of discriminating features which separate the different types of cluster from one another, and which could be used by the user to predict which data sets are likely to achieve the goal.<>