Polyanalyst data analysis technique and its specialization for processing data organized as a set of attribute values

Abstract

The data analysis techniques of the Poly Analyst data mining system [2] are based on the automated synthesis of functional programs treated as the multi-dimensional non-linear regression models. This approach provides the system with two valuable properties: 1) it can discover in data the hidden relations that might be of a great variety of forms, 2) it can explore arbitrarily complexly structured data if the corresponding data access primitives are provided. The paper contains a formal description of the final version of the basic Poly Analyst mechanisms, which are utilized in the general case, as well as in a particular case of data organized as a set of attribute values (SAV), which is the most common format for data explored by KDD methods.