Identifikasi Jenis Shorea Menggunakan Jaringan Syaraf Tiruan Propagasi Balik Resilient Berdasarkan Karakteristik Morfologi Daun

Abstract

Shorea (Meranti) is one of the best kind of wood from tropical forests in Indonesia, and one of the best timber of the family Dipterocarpaceae. Shorea has the highest species diversity among its family, which consists about 194 species. This makes it difficult to identify the species of Shorea. Experts in the field of automatic Shorea identification is still rare. Hence, identification using the morphological characteristic of Shorea leaves is more preferred. This research is focused to identify Shorea species using morphological characteristics of leaves. The data used are manual measurement from a sample of the leaves which are taken from each type of Shorea ovalis, Shorea leprosula, Shorea platyclados, Shorea seminis, and Shorea beccariana. Sample of the leaves is taken from the Shorea collection at the Bogor Botanical Gardens. Classification model is built by using resilient backpropagation neural network. After the data has been collected, the feature values of these five species of Shorea is used in the experiment for training and testing the resilient backpropagation neural network. In the training process, we use a variation of the resilient backpropagation neural network parameters in a 5-fold cross-validation experiment. Each experiment unit is divided into two types: with regression and without regression. Overall, the accuracies of the experiment was 98%. Both types produce the same accuracy with different resilient backpropagation neural network optimal parameters, and the same incorrect identification, which is an instance of Shorea leprosula being identified as an instance of Shorea ovalis.