Please use this identifier to cite or link to this item: http://hdl.handle.net/10609/148753
Title: Detección de fraude alimentario en leche: Análisis de especiación de leche y detección de leche de cabra adulterada con leches de menor calidad, empleando aprendizaje automático e implementación en aplicación web
Other Titles: Detecció de frau alimentari en llet: Anàlisi d’especiació de llet i detecció de llet de cabra adulterada amb llets de qualitat menor, emprant aprenentatge automàtic i implementació en aplicació web
Food fraud detection in milk: Milk speciation analisys and detection of adulterated goat milk with lower quality milk, using machine learning and implementation in web-based aplication
Author: López González, Miguel Ángel
Tutor: Rebrij, Romina  
Others: Ventrua Royo, Carles
Abstract: Food fraud is a risk that compromises the quality and safety of food and involves an economic disadvantage. This work carries out a study of speciation and adulteration of milk from mass spectrometry data. It aims to find the most efficient models and implement them into a tool that facilitate the detection of fraud in milk. The following algorithms have been used: support vector machines (SVM), artificial neural networks (NN) and random forests (RF), from which several models with different parameters where designed. High dimensionality of data and scarcity of samples involved sample increasing using the SMOTE technique and also involved dimension reduction by principal component analysis (PCA). The application of these techniques together and separately has generated four working scenarios. The determination of the best model was based on accuracy, Kappa, sensitivity, and specificity metrics, as well as the prioritization of simplest model. SVM models with linear kernel were chosen as optimal, achieving 100% accuracy in speciation and 96.3% in adulteration. The ability of selected models to detect fraud in milk with a minimum accuracy of 90% has been demonstrated. It is also demonstrated that reducing dimensions and increasing data is the best scenario, improving efficiency. Therefore, it was found that PCA and SMOTE are good techniques for this aim. Finally, the selected models were implemented in a web-based tool, making accessible milk fraud checking in a simple and fast way.
Keywords: machine learning
food fraud
food adulteration
milk
food safety
support vector machines
artificial neural networks
random forests
bionformatics
biostatistics
Document type: info:eu-repo/semantics/masterThesis
Issue Date: 20-Jun-2023
Publication license: http://creativecommons.org/licenses/by-nc-sa/3.0/es/  
Appears in Collections:Trabajos finales de carrera, trabajos de investigación, etc.

Files in This Item:
File Description SizeFormat 
malopezgonzalezTFM0623memoria.pdfMemoria del TFM1,92 MBAdobe PDFThumbnail
View/Open
Share:
Export:
View statistics

This item is licensed under aCreative Commons License Creative Commons