Potential methods in pattern recognition

Geometrical bounding of data space and nonlinear classification of chemical data using MPGA algorithm

This paper develops a multi-parturition genetic algorithm (MPGA) to be used in geometrical bounding of the overlapped clusters in a data set for the classification of chemical data. Two new operators have been introduced to modify the conventional genetic algorithm, namely, multi-parturition and decimation and orientated creation to improve the linear classification results and diminish the computational time. To circumvent the difficulty commonly encountered in the treatment of linearly inseparable chemical data sets, the optimized linear classifier is further modified to provide a complementary nonlinear classifier. For this reason the space regions of the overlapped clusters have been bounded by erection of half-hyperellipsoids over the linearly misclassified patterns. The proposed MPGA was applied to classify a number of chemical and other data sets with a dimension from 4 to 14. Experimental results have indicated that the proposed MPGA could classify seriously overlapped data sets with an acceptable error rate.

The influence of data pre-processing in the pattern recognition of excipients near-infrared spectra

The effect of data pre-processing (no pre-processing, offset correction, de-trending, standard normal variate transformation (SNV), SNV + de-trending, multiplicative scatter correction, first and second derivative transformation after smoothing) on the identification of ten pharmaceutical excipients is investigated. Four pattern recognition methods are tested in the study, namely the Mahalanobis distance method, the SIMCA residual variance method, the wavelength distance method and a method based on triangular potential functions. The performance of the 32 method combinations is evaluated on the basis of two NIR data sets. The first one, measured in 1994, is used to build the classification models, the second, measured from 1994-1997, is used to assess the quality of the models. The best approach for the given data sets is the wavelength distance method combined with de-trending, a simple baseline correction method. More general recommendations for pre-processing excipient NIR data and for choosing an appropriate classification method are given.