Ionomics-based machine learning classification model to discriminate the geographical origin of Philippine carabao mango (Mangifera indica L.)

• Main Author: Laurio, Christian D.
• Format: text
• Language: English
• Published: Animo Repository 2023
• Subjects: Mango > Philippines; Chemistry
• Online Access: https://animorepository.dlsu.edu.ph/etdm_chem/18; https://animorepository.dlsu.edu.ph/context/etdm_chem/article/1018/viewcontent/2023_Laurio_Ionomics_based_machine_learning_classification_model_Full_text.pdf
• Institution: De La Salle University

The Philippine carabao mango is a high-value agricultural product. It is, however, affected by rising concerns about food fraud and origin-mislabeling resulting in consumer distrust, decline in production, and exportation. There is a need for a long-term, strategic research and development initiatives especially using innovative technologies that would help enhance its competitiveness in the local and international markets. We report herein the use of modern technology such as artificial intelligence to classify and authenticate mango samples. In this work, the geographical origin of 70 mango samples collected from Guimaras and Zambales was determined using ionomics technique and machine learning. Seventeen ionomes were analyzed using a validated method of Inductively-coupled Plasma Mass Spectrometry (ICP-MS) and were subjected to multivariate analysis, including correlation, principal component analysis (PCA), and partial least square–discriminant analysis (PLS-DA). PCA and PLS-DA clearly discriminated the samples between two locations. Both variable importance of projection (VIP) scores and F-scores agreed that the ionomes contributed significantly to the origin discrimination identifying the Ni, V, Mn, Ca, Ba, Fe, and Cu ions as chemical markers. These ionomes were used to develop machine learning classification models namely: Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) or Multi-layer Perceptron (MLP). The prediction accuracy of RF, SVM, and ANN/MLP models reached 87.5%, 100%, and 100%, respectively, allowing for the reliable authentication of mango origin from Guimaras and Zambales. This study contributes to the body of knowledge about the applications of both ionomics and machine learning for the determination of fruit’s geographical traceability, which can be used for controlling the geographical origin of mango by the government authorities and protecting consumers from improper labeling and unfair trade.