RMSEP development

The full-scale industrial experiment demonstrated the feasibility of a convenient, nonintrusive aconstic chemometric facility for reliable ammonia concentration prediction. The training experimental design spanned the industrial concentration range of interest (0-8%). Two-segment cross-validation (test set switch) showed good accnracy (slope 0.96) combined with a satisfactory RMSEP. It is fully possible to further develop this pilot study calibration basis nntil a fnll industrial model has been achieved. There wonld appear to be several types of analogous chemical analytes in other process technological contexts, which may be similarly approached by acoustic chemometrics. 

NIR models are validated in order to ensure quality in the analytical results obtained in applying the method developed to samples independent of those used in the calibration process. Although constructing the model involves the use of validation techniques that allow some basic characteristics of the model to be established, a set of samples not employed in the calibration process is required for prediction in order to conhrm the goodness of the model. Such samples can be selected from the initial set, and should possess the same properties as those in the calibration set. The quality of the results is assessed in terms of parameters such as the relative standard error of prediction (RSEP) or the root mean square error of prediction (RMSEP). 

RMSEP = [L(y - J rcfcrcncc) /A] (A = numbcr of validation samples), with concentration units. The behaviour of RMSEP is depicted at Figure 4.14, i.e. it is high whenever a too low or too high number of latent variables are included in the model and it decreases more or less sharply at the vicinity of the optimum number of factors. Nevertheless, you should not use the validation set to select the model because, then, you would need another true validation set. As was explained in previous sections, if you have many samples (which is seldom the case) you can develop three sample sets one for calibration, one for fine-tuning the model and, finally, another one for a true validation. 

Regardless of potential collection efficiency improvements, they obtained an RMSEP of 0.38 mM. However, the PLS calibration model was obtained using 30 samples with 12 factors retained for the development of the regression vector. Without reported evidence of glucose spectral features, it is difficult to determine whether the data were overfit. The model was applied to 24 samples that were not in the calibration set, giving some justification to the analysis. 

The developed models should be tested using independent samples as validation sets to verify model accuracy and robustness. To evaluate model accuracy, the statistics used were the coefficient of correlation in calibration (rc i), coefficient of correlation in prediction (rpred), root mean square error of calibration (RMSEC), and root mean square error of prediction (RMSEP). 

This conclusion was obtained after developing a predictive model similar to the model presented in (2), but using calibration samples that included varying amounts of highly oxidized MEA (polluted samples). The RMSEP of the calibrated model, being 0.46 and 0.23 wt% for MEA and CO2 respectively, is somewhat higher than for the unpolluted model of (2). However, when applying the polluted model to the experimental data discussed in section 3.1, an improved predictive accuracy is obtained 0.57... 

The validation spectra are input to the net, the concentrations are predicted and the predictions are compared to the target concentrations. Of course, neither the weights nor the biases are modified at this stage. The objective is to develop models that predict the validation samples with minimum error. In PLS this was the RMSEP (root mean square error of prediction), and although of course it can also be used here, the MSE (mean square error) is used more... 

---