Learning Vector Quantization for Multiclass Classification:  Application to Characterization of Plastics

Lloyd, Gavin R.; Brereton, Richard G.; Faria, Rita; Duncan, John C.

doi:10.1021/ci700019q

Cited by 33 publications

(17 citation statements)

References 35 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This information is shown in Figure 3(a) and Figure 3(b). Classification using [5], the clustering of the fabric materials on the PCA plot can be related to chemical properties of the fabric materials. All characteristic group frequencies for organic functional groups were obtained from Principles of Instrumental Analysis, 7th edition by Skoog, Holler and Crouch [12].…”

Section: Correlation-pcamentioning

confidence: 99%

Analysis of ATR-FTIR Absorption-Reflection Data from 13 Polymeric Fabric Materials Using Chemometrics

Pumure¹,

Ford²,

Shannon³

et al. 2015

AJAC

View full text Add to dashboard Cite

We used both correlation and covariance-principal component analysis (PCA) to classify the same absorption-reflectance data collected from 13 different polymeric fabric materials that was obtained using Attenuated Total Reflectance-Fourier Transform Infrared spectroscopy (ATR-FTIR). The application of the two techniques, though similar, yielded results that represent different chemical properties of the polymeric substances. On one hand, correlation-PCA enabled the classification of the fabric materials according to the organic functional groups of their repeating monomer units. On the other hand, covariance-PCA was used to classify the fabric materials primarily according to their origins; natural (animal or plant) or synthetic. Hence besides major chemical functional groups of the repeat units, it appears covariance-PCA is also sensitive to other characteristic chemical (inorganic and/or organic) or biochemical material inclusions that are found in different samples. We therefore recommend the application of both covariance-PCA and correlation-PCA on datasets, whenever applicable, to enable a broader classification of spectroscopic information through data mining and exploration.

show abstract

Section: Correlation-pcamentioning

confidence: 99%

Analysis of ATR-FTIR Absorption-Reflection Data from 13 Polymeric Fabric Materials Using Chemometrics

Pumure¹,

Ford²,

Shannon³

et al. 2015

AJAC

View full text Add to dashboard Cite

show abstract

“…Table 3 show the overall effectiveness of the classification models for SVM [12], LVQ [13], LDA [14], QDA [11] and EDC [11] for %CC of three kind selected data pre-processing method. Evaluated by percentage correctly classified (%CC) was functioned for recognizing which method of variable selection was accepted in developing the classification model.According to Table 3 all five classifier give high value of percentage correctly classified except QDA.…”

Section: Svmmentioning

confidence: 99%

Data preprocessing methods of FT-NIR spectral data for the classification cooking oil

Rasaruddin¹,

Fong²,

Jaafar³

2014

AIP Conference Proceedings

View full text Add to dashboard Cite

This recent work describes the data pre-processing method of FT-NIR spectroscopy datasets of cooking oil and its quality parameters with chemometrics method. Pre-processing of near-infrared (NIR) spectral data has become an integral part of chemometrics modelling. Hence, this work is dedicated to investigate the utility and effectiveness of preprocessing algorithms namely row scaling, column scaling and single scaling process with Standard Normal Variate (SNV). The combinations of these scaling methods have impact on exploratory analysis and classification via Principle Component Analysis plot (PCA). The samples were divided into palm oil and non-palm cooking oil. The classification model was build using FT-NIR cooking oil spectra datasets in absorbance mode at the range of 4000cm -1 -14000cm -1 . Savitzky Golay derivative was applied before developing the classification model. Then, the data was separated into two sets which were training set and test set by using Duplex method. The number of each class was kept equal to 2/3 of the class that has the minimum number of sample. Then, the sample was employed t-statistic as variable selection method in order to select which variable is significant towards the classification models. The evaluation of data pre-processing were looking at value of modified silhouette width (mSW), PCA and also Percentage Correctly Classified (%CC). The results show that different data processing strategies resulting to substantial amount of model performances quality. The effects of several data pre-processing i.e. row scaling, column standardisation and single scaling process with Standard Normal Variate indicated by mSW and %CC. At two PCs model, all five classifier gave high %CC except Quadratic Distance Analysis.

show abstract

“…For this analysis the Loss Modulus (E 00 ) was used, which represents the energy dissipated within the sample per oscillation. Full experimental details and other work on this data have been described previously [47][48][49][50][51]. The temperature range in this study was from À518C until the minimum stiffness of the polymer was reached, after which no further meaningful data could be collected and the DMA instrument stopped automatically.…”

Section: Dynamic Mechanical Analysismentioning

confidence: 99%

“…The method is illustrated using the thermal profiles of nine groups of polymers studied by Dynamic Mechanical Analysis [45][46][47][48][49][50][51] representing a complex problem where there are nine classes. The structure of each class is often quite different, as each group of polymers consists of several grades or subgroups, hence is not homogeneous.…”

Section: Introductionmentioning

confidence: 99%

Disjoint hard models for classification

Lloyd

Duncan³

et al. 2010

Journal of Chemometrics

View full text Add to dashboard Cite

The paper describes a new approach for disjoint hard modelling of classes. This involves developing independent PC models for each group in the class, and calculating both the Q statistic (square prediction error) for each sample to the class model and a separate statistic about how well samples are classified within the projected PC space. The latter statistic can be applied to different types of classifiers, in this paper we choose to illustrate by Quadratic Discriminant Analysis (D statistic) and one class Support Vector Domain Description (SVDD) ( f-value). The two measures (Q and the classifier dependent statistic) are combined into a joint decision function which uniquely classifies each sample. The disjoint hard models are contrasted to conjoint models where PCA is performed on the entire dataset using both QDA and Support Vector Machines (SVMs) classifiers. The optimum number of PCs for each model is determined using the bootstrap, and model performance assessed on 100 test sets obtained using different iterative splits, using %PA (Predictive Ability) and %CR (Classification Rate). The method is illustrated using a dataset consisting of 293 samples from nine groups of polymers obtained using thermal profiling. The approach described, in this paper, has many of the advantages of one class disjoint models (e.g. SIMCA) and of conventional hard models, and is useful if it is known that all samples must belong to one of a series of known groups but each group has a very different structure.

show abstract

Learning Vector Quantization for Multiclass Classification: Application to Characterization of Plastics

Cited by 33 publications

References 35 publications

Analysis of ATR-FTIR Absorption-Reflection Data from 13 Polymeric Fabric Materials Using Chemometrics

Analysis of ATR-FTIR Absorption-Reflection Data from 13 Polymeric Fabric Materials Using Chemometrics

Data preprocessing methods of FT-NIR spectral data for the classification cooking oil

Disjoint hard models for classification

Contact Info

Product

Resources

About