Multi-elemental analysis and geographic identification of Philippine coffee beans from Luzon (Regions I, II, and CAR) using x-ray fluorescence spectroscopy and machine learning methods
Date of Publication
11-2025
Document Type
Bachelor's Thesis
Degree Name
Bachelor of Science in Chemistry Minor in Business Studies
Subject Categories
Chemistry
College
College of Science
Department/Unit
Chemistry
Thesis Advisor
Emmanuel V. Garcia
Angel T. Bautista, VII
Defense Panel Chair
Lourdes P. Guidote
Defense Panel Member
Mariafe Calingacion
Aldrin P. Bonto
Abstract (English)
Coffee quality and authenticity are strongly affected by many factors including its geographical origin, environmental conditions, and elemental makeup. This study aims to establish reliable scientific methods in order to verify and trace Philippine coffee in local and international market. This study performed multi-elemental profiling of Philippine coffee beans from Luzon and evaluated classification by region (Regions 1, 2, CAR) and variety (Arabica vs Robusta). Forty-nine (49) samples were pulverized, sieved, and pelletized for X-ray Fluorescence (XRF) analysis. After screening, 14 elements were retained (ranked by mean decreasing concentration: K > Mg > P > S > Cl > Al > Cr > Mn > Sr > Pd > Cu > Rb > Ni > Zn). Data were analyzed using one-way omnibus tests and machine learning methods, specifically principal component analysis (PCA), linear discriminant analysis (LDA), and random forest (RF). Unsupervised PCA revealed chemically meaningful gradients: CAR associated with higher Mn, Mg, Zn, S; Region 1 with higher K, P, Cu; and Region 2 with higher Cl and lower Rb. Patterns were consistent with one-way omnibus tests that identified significant differences in Cl, S, Rb, Cu, K, Mg, Mn, and Zn. For supervised models, LDA achieved 72.9% LOOCV accuracy, while Random Forest (RF) reached 81.3% OOB accuracy for regional classification. In contrast, varietal classification was markedly stronger with 93.9% LOOCV accuracy for LDA, and 97.7% OOB accuracy for RF. Overall, p-XRF based multi-elemental analysis combined with machine learning methods provides regional and varietal discrimination, but variety exerts a more systematic influence than region on elemental profiles of these coffee samples.
Abstract Format
html
Abstract (Filipino)
Coffee quality and authenticity are strongly affected by many factors including its geographical origin, environmental conditions, and elemental makeup. This study aims to establish reliable scientific methods in order to verify and trace Philippine coffee in local and international market. This study performed multi-elemental profiling of Philippine coffee beans from Luzon and evaluated classification by region (Regions 1, 2, CAR) and variety (Arabica vs Robusta). Forty-nine (49) samples were pulverized, sieved, and pelletized for X-ray Fluorescence (XRF) analysis. After screening, 14 elements were retained (ranked by mean decreasing concentration: K > Mg > P > S > Cl > Al > Cr > Mn > Sr > Pd > Cu > Rb > Ni > Zn). Data were analyzed using one-way omnibus tests and machine learning methods, specifically principal component analysis (PCA), linear discriminant analysis (LDA), and random forest (RF). Unsupervised PCA revealed chemically meaningful gradients: CAR associated with higher Mn, Mg, Zn, S; Region 1 with higher K, P, Cu; and Region 2 with higher Cl and lower Rb. Patterns were consistent with one-way omnibus tests that identified significant differences in Cl, S, Rb, Cu, K, Mg, Mn, and Zn. For supervised models, LDA achieved 72.9% LOOCV accuracy, while Random Forest (RF) reached 81.3% OOB accuracy for regional classification. In contrast, varietal classification was markedly stronger with 93.9% LOOCV accuracy for LDA, and 97.7% OOB accuracy for RF. Overall, p-XRF based multi-elemental analysis combined with machine learning methods provides regional and varietal discrimination, but variety exerts a more systematic influence than region on elemental profiles of these coffee samples.
Abstract Format
html
Language
English
Format
Electronic
Keywords
Coffee--Philippines; X-ray spectroscopy
Recommended Citation
De Vera, N. D., & Yap, A. E. (2025). Multi-elemental analysis and geographic identification of Philippine coffee beans from Luzon (Regions I, II, and CAR) using x-ray fluorescence spectroscopy and machine learning methods. Retrieved from https://animorepository.dlsu.edu.ph/etdb_chem/73
Upload Full Text
wf_yes
Embargo Period
1-31-2026
