Determining the Geographical Origin of Crude Palm Oil with the Combined Use of GC-IMS Fingerprinting and Chemometrics
Current administrative controls used to verify geographical provenance within palm oil supply chains require enhancement and strengthening by more robust analytical methods. In this study, the application of volatile organic compound fingerprinting, in combination with five different analytical classification models, has been used to verify the regional geographical provenance of crude palm oil samples. For this purpose, 108 crude palm oil samples were collected from two regions within Malaysia, namely Peninsular Malaysia (32) and Sabah (76). Samples were analysed by gas chromatography-ion mobility spectrometry (GC-IMS) and the five predictive models (Sparse Logistic Regression, Random Forests, Gaussian Processes, Support Vector Machines, and Artificial Neural Networks) were built and applied. Models were validated using 10-fold cross-validation. The Area Under Curve (AUC) measure was used as a summary indicator of the performance of each classifier. All models performed well (AUC 0.96) with the Sparse Logistic Regression model giving best performance (AUC = 0.98). This demonstrates that the verification of the geographical origin of crude palm oil is feasible by volatile organic compound fingerprinting, using GC-IMS supported by chemometric analysis.