Intelligent odour profile classification of honey using Case Based Reasoning and K-Nearest Neighbors /
Nurdiyana Zahed,
Intelligent odour profile classification of honey using Case Based Reasoning and K-Nearest Neighbors / Nurdiyana Zahed - xii, 102 pages : illustrations (some color) ; 30 cm. + 1 CD ROM
Faculty of Electrical and Electronics Engineering
Thesis (Master of Science) -- Universiti Malaysia Pahang – 2018
Includes bibliographical references
Honey produced by sting bee contains several important nutrients that can be used as food additive, medicinal food and food preservative. In Malaysia, three most common types of pure honey are Gelam, Acacia, and Tualang which can be verified from Department of Agriculture Malaysia. Honey can be classified using several methods and one of them is by referring to its botanical origin (botanical place where the honey is collected) which influence the quality of honey, its market price and benefits. Various techniques can be found in classifying honey where odour profile is amongst them. In order to establish honey odour profile, electronic nose (E-nose) is one of the significant instrument that can be used for honey odour profile measurement and detection. E-nose comprises of three major components; chamber, microcontroller and sensor arrays that capable of establishing odour profile pattern. In order to get E-nose response measurement, it has been validated using Gas Chromatography Mass Spectrometry (GC-MS) which was pre-processed using Solid Phase Micro-extraction (SPME). Three datasets of three different groups of honey from E-nose were measured in electrical signal for odour detection. The measured resistance data was pre-processed using normalization technique to standardize the scale of sample features. Mean values were extracted as selected feature and statistical methods using proximity matrix and boxplot data pattern graphical representation according to types of honey. Mean features for Case-based Reasoning (CBR) cases that have been extracted were employed into two types of classifiers; intelligent classification using CBR and K-Nearest Neighbors (KNN). Both CBR and KNN performances have been evaluated. The results have shown 100% rate of accuracy, sensitivity and specificity of classification from CBR classifier and KNN using weigh (k=1), ratio 90:10 and Euclidean distance.
THE0008051(Local)
Faculty of Electrical and Electronics Engineering--Dissertations
Universities and colleges--Disertations
Theses
Intelligent odour profile classification of honey using Case Based Reasoning and K-Nearest Neighbors / Nurdiyana Zahed - xii, 102 pages : illustrations (some color) ; 30 cm. + 1 CD ROM
Faculty of Electrical and Electronics Engineering
Thesis (Master of Science) -- Universiti Malaysia Pahang – 2018
Includes bibliographical references
Honey produced by sting bee contains several important nutrients that can be used as food additive, medicinal food and food preservative. In Malaysia, three most common types of pure honey are Gelam, Acacia, and Tualang which can be verified from Department of Agriculture Malaysia. Honey can be classified using several methods and one of them is by referring to its botanical origin (botanical place where the honey is collected) which influence the quality of honey, its market price and benefits. Various techniques can be found in classifying honey where odour profile is amongst them. In order to establish honey odour profile, electronic nose (E-nose) is one of the significant instrument that can be used for honey odour profile measurement and detection. E-nose comprises of three major components; chamber, microcontroller and sensor arrays that capable of establishing odour profile pattern. In order to get E-nose response measurement, it has been validated using Gas Chromatography Mass Spectrometry (GC-MS) which was pre-processed using Solid Phase Micro-extraction (SPME). Three datasets of three different groups of honey from E-nose were measured in electrical signal for odour detection. The measured resistance data was pre-processed using normalization technique to standardize the scale of sample features. Mean values were extracted as selected feature and statistical methods using proximity matrix and boxplot data pattern graphical representation according to types of honey. Mean features for Case-based Reasoning (CBR) cases that have been extracted were employed into two types of classifiers; intelligent classification using CBR and K-Nearest Neighbors (KNN). Both CBR and KNN performances have been evaluated. The results have shown 100% rate of accuracy, sensitivity and specificity of classification from CBR classifier and KNN using weigh (k=1), ratio 90:10 and Euclidean distance.
THE0008051(Local)
Faculty of Electrical and Electronics Engineering--Dissertations
Universities and colleges--Disertations
Theses