Towards authentication of beef, chicken and lard using micro near-infrared spectrometer based on support vector machine classification
NIR (near infrared) spectrometer utilized a quick reliable mean of molecular chemical detection. In this paper, we propose a method on authenticating fats originated from beef, chicken and lard. These compositions can be identified by NIR spectrometers through qualitative and quantitative analysis...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English English |
| Published: |
Asian Research Publishing Network (ARPN)
2016
|
| Subjects: | |
| Online Access: | http://irep.iium.edu.my/51090/ http://irep.iium.edu.my/51090/ http://irep.iium.edu.my/51090/1/jeas_0316_3933.pdf http://irep.iium.edu.my/51090/4/51090_Towards%20authentication%20of%20beef%2C%20chicken%20and%20lard%20using%20micro%20near%C2%ADinfrared_Scopus.pdf |
| Summary: | NIR (near infrared) spectrometer utilized a quick reliable mean of molecular chemical detection. In this paper, we
propose a method on authenticating fats originated from beef, chicken and lard. These compositions can be identified by
NIR spectrometers through qualitative and quantitative analysis. Yet most of the analysis lack the capacity to find a pattern
in the spectrums to be used in classification or regression models. The disadvantage of spectrum after all is the inability to
show the concentration of fatty acids, yet these fatty acid components are shared by all kinds of fat/oil. Therefore, a new
method is proposed to create a clear and a distinguishable pattern for the classification. The spectrum of each group (beef
fat, chicken fat and pig fat “lard”) of samples were acquired using a readymade micro-NIR spectrometer. The raw data
required further processing before using it in the classifier. These processes including standard normal variant and Savitsky-Golay smoothing. Furthermore, the processed data was classified using support vector machine (SVM) with polynomial kernel. The trained SVM model showed 98.33% accuracy for 10-fold cross validation and 86.67% accuracy for unseen/testing data. For each individual kind of fat the model was able to identify the unseen data satisfactorily as follows lard with 100% accuracy and combined, chicken and beef showed 80% accuracy. |
|---|