Electronic-nose modelling and data analysis using a self-organizing map

Abstract

The self-organizing map (SOM) is among the most widely studied and applied types of neural networks; nevertheless, it has not been utilized adequately to model and analyse data of multisensor systems, in particular of chemical sensor arrays. In this paper an example of how electronic noses can take advantage of the utilization of the self-organizing map is illustrated and discussed. In particular a number of ways of extracting information from a trained SOM is presented. The methodology outlined here is valid for any kind of multisensor application, also in fields distant from the world of chemical sensors.