Interactive Feature Extraction From Multivariate Production Data
S. Rezvani, Girijesh Prasad, J. Muir, K. McCraken · Pattern Recognition in Information Systems · 2003
As part of a research programme between the University of Ulsterand a polymeric manufacturing company, the interrelationships andperformance attributes of polymeric components used in adhesivesmanufactured for medical packaging are being studied. To understand theinteractions between the parameters, to select the optimal manufacturingprocesses and to improve economic factors within this investigation, a series ofmultidimensional data processing and visualisation methods such as PrincipalComponent Mapping (PCM), Curvilinear Component Mapping (CCM) andSammon Mapping have been examined. As a novel approach to multivariatedata visualisation and analysis, these techniques were employed in such a waythat interactive multi-layer maps could be generated. Each layer within thisfeature extraction technique corresponds to a specific attribute andcharacteristic of the dataset. Activation and superimpositions of individuallayers allow studying manufacturing attributes, parameter interactions andproduct formulations,