Sistem Pakar Identifikasi Varietas Ikan Mas (Cyprinus carpio) Berdasarkan Karakteristik Morfologi dan Tingkah Laku

Abstract

This research is aimed to identify 6 common carp varieties, using 8 parameters input of morphological characteristics and behavior. K-fold cross validation is used to determinate training data and test data. Nominal variables are processed by the nominal distance, while the numeric and ordinal variables are processed by using Euclidean distance. Before calculating the Euclidean distance, min-max normalization method was applied on the numeric and ordinal variables. The calculation result between Euclidean distance and nominal distance were combined by using the aggregate formula. The classification method used is the k-nearest neighbor (KNN) method. The best average accuracy obtained without normalization was 94.58% and with normalization was 98.54% when k = 3. Based on the expert validation, the system gives a 100% accurac