Kaedah Kolorimetri untuk Analisis Kuantitatif Kapsaisin Secara Pencaman Corak Menggunakan Jaringan Neural Timan

A quantitative study for capsaicin based on the use of 2,6-dichloro-pbenzoquinone-4-chlorimide (Gibbs reagent) and artificial neural network (ANN) has been carried out. The characterization include pH optimization, effect of reagent concentration, dynamic range of capsaicin concentration, photo stabil ity, limit of detection and reproducibility. The optimum response was obtained at pH 11.0 and Gibbs reagent concentration of 2.96 x 1Q4 M. The reproducibility of the method was very satisfactory with RSD values of 3.55 %, 2.44% and 4.52% for capsaicin concentration of 200 ppm, 500 ppm and 800 ppm, respectively. Photostabili ty test showed that the reagent was very stab le with RSD value of0.013% for the duration of 38 hours. A three layer feed-forward neural network was used and network training was performed by using back propagation algorithm. For the determination of capsaicin, a neural network with 20 hidden neurons, 0.00001 % learning rate and trained over 47,738 cycles produced the best result. This network was able to extend the narrow dynamic range of capsaicin from 0 - 200 ppm to 0-600 ppm. The average interpolation error produced by this network was approximately 0.06 %.