Prediction of clothing comfort sensation of an undershirt using artificial neural networks with psychophysiological responses as input data

The clothing comfort sensation is a combination of complex components, including psychological and physiological responses. General linear analysis is not always sufficient for the evaluation of the clothing comfort sensation. The current study sought to predict the clothing comfort sensation of wearing an undershirt using an artificial neural network (ANN). We constructed ANN models with psychological sensation data and physiological response data as inputs, including electrocardiogram and thermo-physiological indicators, and the clothing comfort sensation as the output. For the input layer of the model, three conditions were used: the psychological response data only, the physiological response data only, and both the psychological and physiological data. The number of hidden layers in the models ranged from one to three, and the number of units in each hidden layer was changed when fixed values of 30, 60, and 90 were used, or according to the number of data points in the input conditions. The results revealed that, among the three conditions, the accuracy rate was higher when both psychological and physiological response data were used as input. The prediction results exhibited an accuracy rate of up to 85% for unknown test data. The results suggest that the method of evaluating the state of clothing comfort sensation when wearing an undershirt using psychophysiological response measurement was effective and that neural networks are useful for predicting the clothing comfort sensation.