Abstract
Sensor arrays based on chemical sensors produce multidimensional patterns of data that may be used discriminate between different chemicals. For the human observer, visualization of multidimensional data is difficult, since the eye and brain process visual information in two or three dimensions. To devise a simple means of data inspection from the response of sensor arrays, PCA (Principal Component Analysis) or Sammon's nonlinear mapping technique can be applied. The PCA, which is a well-known statistical method and widely used in data analysis, has disadvantages including data distortion and the axes for plotting the dimensionally reduced data have no physical meaning in terms of how different one cluster is from another. In this paper, we have investigated two techniques and proposed a combination technique of PCA and nonlinear Sammom mapping for visualization of multidimensional patterns to two dimensions using data sets from odor sensing system. We conclude the combination technique has shown more advantages comparing with the PCA and Sammon nonlinear technique individually.