Chemometrics method for classification of cooking oils from FT-MIR and FT-NIR spectra / Mas Ezatul Nadia Mohd Ruah
Traditional analyses for cooking oil authentication which are based on chemical and physical methods have several drawbacks such as slow result outcome, necessity for pre-treatments, the needs for highly skilled personnel to handle instrument and test samples destruction. A combination of Fourier Tr...
Saved in:
| Main Author: | |
|---|---|
| Format: | Thesis |
| Language: | English |
| Published: |
2017
|
| Subjects: | |
| Online Access: | https://ir.uitm.edu.my/id/eprint/79315/1/79315.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Traditional analyses for cooking oil authentication which are based on chemical and physical methods have several drawbacks such as slow result outcome, necessity for pre-treatments, the needs for highly skilled personnel to handle instrument and test samples destruction. A combination of Fourier Transform Mid Infrared (FT-MIR) and Near Infrared (FT-NIR) spectroscopy with chemometrics has proven to be a successful analytical method for variety\of food products. These techniques particularly may possess certain advantages such as rapid measurement, moderate instrument cost and relative ease of sample preparation. The aims of this study are to classify cooking oil into two batches of group which are Batch 1 (palm oil and non palm oil) and Batch 2 (unused palm oil and used palm oil). The samples were analyzed by FT-MIR and FT-NIR spectroscopy and processed using classification methods; linear discriminant analysis (LDA), learning vector quantization (LVQ) and support vector machines (SVM), quadratic discriminant analysis (QDA), and euclidean distance centroid (EDC). The classification model was built using FT-MIR and FT-NIR cooking oil spectra datasets in absorbance mode in the range of 650 to 4000 cm'1 and 4000 - 14000 cm"1 , respectively. Savitzky Golay derivative was applied before developing the classification model. Then, the data was separated into two sets (training set and test set) by using Duplex method and 100 split random methods. The best variable selection method known as ^-statistic was applied to the datasets in order to find the most significant variable. Column standardisation is the best data pre processing method for both spectroscopy methods. The evaluation of data pre processing was evaluated by using modified silhouette width (mSW). Then, it was followed with finding the value of percentage correctly classified (%CC) for every model in order to show the performance of developed classification models. The combination of FT-MIR and FT-NIR spectroscopy with chemometrics method showed the ability of classifying the sample into the interest groups of sample which are palm oil, non-palm oil, used palm oil and unused palm oil with using 2 principal components (PC's). |
|---|