concentrate products (see Fig. 1a–d) the value of two informational
peaks on 230/310 and 285/310 nm decreases in relation to the
peaks on 260/340 and 320/430 nm.
The task for application of analysis for ISC analyser is rapid measurement
and classification of commercial apple juice type. The
k-Nearest Neighbor (kNN) Search technique is used in classification
procedure (Shakhnarovish, Darrell, & Indyk, 2006). Minkowski
metric of second order is selected as a distance measure for kNN.
The procedure of analysis is defined in steps. First, new sample
SFS is embedded into two dimensional PCs space using transform
matrix obtained from statistical analysis. Then kNN classifier is
constructed in PCs space using reference dataset of 59 samples
for learning. At last the nearest neighbourhood is found for
instance and sample is classified. The separate novelty detection
algorithm is used to exclude samples that do not apply to the apple
products