| • Preprocess spectral/chromatographic data (baseline, smoothing, derivatives, normalization, SNV/MSC, alignment) for robust modeling (k2, k3, k4).
• Perform and interpret exploratory analysis (PCA, clustering; scores, loadings, outliers) (k2, k3, k4, k5).
• Build and validate calibration models (PCR/PLS) using cross validation; report RMSE, R2, and bias (k3, k4, k5).
• Develop and assess classification models (LDA, kNN, SVM/PLS DA) with confusion matrices and ROC/AUC (k3, k4, k5).
• Plan and analyze simple experiments (screening/factorial, response surfaces) (k2, k3, k4).
• Produce reproducible analyses in R/Python/MATLAB and communicate results in a scientific style (k3, k4, k5).
|
The lecture covers:
• Chemometric foundations: data structures, measurement error, basic statistical tests, overfitting vs. generalization.
• Data preprocessing for spectra/chromatograms: detrending, Savitzky–Golay, derivatives, scaling, SNV/MSC, retention time correction.
• Exploratory analysis: PCA and clustering; leverage/outlier diagnostics; cluster analysis.
• Calibration and classification: MLR, PCR, PLS; LDA/kNN/SVM/PLS DA; validation strategies and key figures of merit (RMSE, R2, sensitivity/specificity).
• Design of experiments (DoE): screening, factorials, response surface methods;
• Process monitoring (MSPC) and PAT basics, measure uncertainty: Hotelling’s T2, SPE/Q control limits., etc.
|
