Date of Publication


Document Type

Bachelor's Thesis

Degree Name

Bachelor of Science in Biochemistry

Subject Categories



College of Science



Thesis Advisor

Emmanuel V. Garcia

Defense Panel Member

Marissa G. Noel
Joan Candice Ondevilla


The authenticity and the premium associated with specialty coffee cultivated in specific coffee-growing regions in the Philippines increases its value and demand, making it vulnerable to adulteration and counterfeiting. As such, the chemical analysis of coffee beans for the purpose of establishing the authenticity of coffee beans in terms of variety and provenance becomes imperative. In the study, 11 samples of green coffee beans (Arabica, Robusta, Excelsa, and Liberica) harvested from different regions in the Philippines (Cordillera Administrative Region, CALABARZON, Western Visayas, Central Visayas and Caraga) were subjected to multi-elemental analysis using X-ray Fluorescence spectrometry (XRF). The analysis of the samples was done in triplicates. In the analysis, a total of 21 elements were detected in the green coffee bean samples: Mg, Al, P, S, Cl, K, Cr, Mn, Ni, Cu, Zn, As, Rb, Sr, Y, Nb, Pd, W, Pt, Bi, and U. Among the elements detected, potassium (K), magnesium (Mg), and sulfur (S) were found to have the highest average concentrations (%wt) in all samples. Contrary to this, arsenic (As), bismuth (Bi), and Yttrium (Y) were found to have low average concentrations (%wt) in the samples. The four varieties of coffee were revealed to have distinct elements that would allow differentiation based on the dominant elemental composition. Likewise, the results also showed distinguishing elemental profile characteristics across the green coffee bean samples from the five sampling regions. A machine learning technique, specifically Random Forest, was used to generate a classification model for the prediction of the variety and geographical origin of coffee using the multi-elemental data from the XRF analysis. The study demonstrates the potential of the multi-elemental profile of coffee beans as effective discriminants and be used as elemental fingerprints for the identification of coffee variety and the establishment of coffee provenance. Overall, the study shows that XRF-based multi-elemental profiling technique combined with machine learning algorithms (random forest) is a promising tool for coffee authentication and fraud detection.

Keywords: Elemental profiling, Geographical discrimination, Varietal discrimination, Machine Learning, X-ray Fluorescence Spectrometry

Abstract Format






Physical Description

xii, 93 leaves


Coffee--Philippines; X-ray spectroscopy; Machine learning

Upload Full Text


Embargo Period