Fiabilidad de clasificación con PLS discriminante

Identificador:  TDX:820

Handle: https://hdl.handle.net/20.500.11797/TDX820

Autores:  Pérez Pérez, Néstor Fredy

Resumen:
In this thesis the multivariate classification of samples using discriminant partial least squares (p-DPLS) was developed. The method, besides classifying a sample using DPLS, provides the reliability of such classification. p-DPLS was improved to solve multi-class problems and was more suitable than other classification methods such as CART and SIMCA, besides to providing the reliability of classification. Additionally, a methodology for establishing multivariate specification of food commodities was devised. The methodology combined Hotelling T2, SPE and the PLS model predictions. The above developments were applied to classification of soils analyzed by XRF and ICP-MS, honeys analyzed by GC-MS and olive oils analyzed by NIR and 1H-NM1R.