TY  - MANSCPT
AU  - Fathimah Abdul Halim
TI  - Classification of ammonia from water based on odor-profile using K-NN and CBR
SN  - THE0005176(Local)
PY  - 2016///
CY  - Kuantan, Pahang
PB  - UMP
KW  - Faculty of Electrical and Electronics Engineering
KW  - Dissertations
KW  - Universities and Colleges
KW  - Theses
N1  - Faculty of Electrical and Electronics Engineering; Thesis (Doctor of Philosophy (Instrumentation Engineering)) -- Universiti Malaysia Pahang – 2016; Bibliography : p. 92-101
N2  - This project presents the classification of ammonia concentration in water. High concentration of ammonia in the water possess negative effect toward the environment, especially marine’s life. Thus, monitoring and supervising the concentration of ammonia is crucial in order to maintain water quality. Classifying ammonia concentration from the water using electronic nose (E-nose) based on classification technique and signal processing approach is one of the popular method. E-nose consists of four gas sensors where the sensor array are used to produce a unique profile of an odor. The aim of this project is to establish classification method based on odor-profile of ammonia using E-nose. Two groups of ammonia concentration; high (20, 25 ppm) and low (5, 10, 15 ppm) have been measured by using E-nose. Before the ammonia concentration were measured by E-nose, the samples were validated using volatile organic compound (VOC) detector MiniRAE 3000. The confirmed concentrations from VOC detector then were proceeded using E-nose. The data from the E-nose measurements have been pre-processed and normalized using normalization technique in order to obtain ammonia 2D and 3D odor-profile pattern. From the odor-profile pattern, the mean feature was extracted. The extracted feature has been statistically validated using the box plot, proximity matrix and regression analysis. The significant selected normalization and mean features were applied as an input features to k-nearest neighbor (k-NN) and case-based reasoning (CBR) techniques in order to classify the features either as a low or high concentration. The classification results of the k-NN and CBR were evaluated based on performance measures evaluation methods. It was found that the k-NN classifier and the CBR classification techniques were able to classify the different concentration of ammonia from the water. K-NN was used as validation classifier system for CBR. Based on the k-NN and CBR performance measures results, it has been observed that the classification rate of k-NN and CBR are 100%
ER  -