A study on the factor number determination methods in the partial least squares model for the urinalysis using Raman spectroscopy

As an effort for the development of the nonintrusive measurement system, Raman spectroscopy was applied for the urinalysis. By using Raman spectroscopy, the concentration of the urine components could be measured. As a multivariate method, the partial least squares method was performed. When composing a calibration model, the determination of the appropriate number of factors was very important for the accurate prediction of the constituent concentration. In this study, the number of factors was determined by observing the minimum PRESS (prediction residual error sum of squares) value and the biggest correlation coefficient between the predicted values and the original values of the training set. After obtaining the most suitable number of factors by the two methods, the unknown constituent concentration was predicted with those factors, and the correlation coefficients between the pre-examined value and the predicted results from the unknown spectra were calculated. The analysis results using the two factor number determination methods were compared in order to determine which one is more suitable for the accurate urine component concentration prediction. The selected method will be used in the urinalysis using Raman spectroscopy system which will be one of the non-intrusive measurement systems in a live-alone patient´s house.