KOMPARASI METODE FUZZY LVQ DAN RBF PADA KLASIFIKASI KOPI BERBASIS ELECTRONIC NOSE; COMPARISON BETWEEN FUZZY LVQ AND RBF METHOD IN COFFEE CLASSIFICATION BASED ON ELECTRONIC NOSE

Mahardika, Grehasta, Triyogatama Wahyu Widodo

2015 | Skripsi | FMIPA

Classification is one of the method to extract information. The purpose of classification is to analize input data and create the acurate model for each class based on available data. Many method used to do classification, one of them is Artificial Neural Network (ANN). The optimal artificial neural network method to do classification is needed. Radial Basis Function (RBF) is one of artificial neural network that used to solve function approach, control system and classification. The advantage of this method rely on the simple design, good generalization and high noise tolerance in input, however like other multi layer architecture method, it need longer computation times compared to the single layer architecture. Learning Vector Quantization (LVQ) is one of the artificial neural network that use single layer architecture. LVQ method generally used in pattern classification. The advantage of this method are the simplicity of the architecture and short computational time, however according to experiment that has been done, RBF method is better than LVQ. Efforts to optimize LVQ method has been proposed, one of them is combining Fuzzy logic to its process, to obtain a hybrid method called Fuzzy LVQ. Result from this research, which use 48 test data from 6 coffee samples, Fuzzy LVQ method has higher accuracy than RBF method in identifying test data. Accuracy of Fuzzy LVQ is 100% while RBF is 62,5%. The computational time of Fuzzy LVQ is shorter compared to RBF, Fuzzy LVQ is about 1,5 seconds and RBF is about 55,25 seconds.

Kata Kunci : Artificial neural network; gas sensor; hybrid method.