Artificial neural networks compared to factor analysis for low-dimensionalclassification of high-dimensional body fat topography data of healthy anddiabetic subjects

Subcutaneous adipose tissue thickness was measured in 590 healthy subjects at 15 specific body sites by means of the new optical device, lipometer, pr oviding a high-dimensional and partly highly intercorrelated set of data, w hich had been analyzed by factor analysis previously. N-2-N back-propagatio n neural networks are able to perform low-dimensional display of high-dimen sional data as a special application. We report about the performance of su ch a 15-2-15 network and compare its results with the output of factor anal ysis. As test data for verification, measurement values on women with prove n diabetes mellitus type II (NIDDM) are used. Surprisingly our 15-2-15 neur al network is able to reproduce the classification pattern resulting from f actor analysis very precisely. After extracting the network weights the cla ssification of new subjects is even more simple with the neural network as compared with factor analysis. In addition, the network weights are able to cluster highly correlated body sites nicely to different groups, correspon ding to different regions of the human body. Thus, the analysis of these wr ights provides additional information about the structure of the data. Ther efore, N-2-N networks seem to be a good alternative method for analyzing hi gh-dimensional data with strong intercorrelation. (C) 2000 Academic Press.