Abstract: The paper highlighted the problems of selection methods for the preparation of the "raw" data for subsequent processing in the systems of intelligent information analysis. Each application domain contains a lot of different types of identification characteristics. Most algorithms are unable to work directly with all types and formats of data. Therefore, the preparation or the transformation of the input data set is an integral part of analysis system design. In the paper discusses methods of transformation of continuous and discrete types of data useful for analysis of the vector set. On the example of the domain "Information customs control" has been shown as configured training set for recognition of risks violation of customs legislation, based on a neural network the type multilayer perceptron. Were also considered methods of forming pseudographic patterns from the input sequence data initially did not a graphic of origin.

Keywords: preprocessing, transformation data

ACM Classification Keywords: D.2 SOFTWARE ENGINEERING – D.2.12 Interoperability – Data mapping

Link:

PREPROCESSING "RAW" DATA SETS AS AN IMPORTANT ASPECT OF INTELLIGENT INFORMATION PROCESSING

Sergii Konovalenko

http://foibg.com/ibs_isc/ibs-27/ibs-27-p21.pdf