Mushroom Identification using Backpropagation Method
Muhammad Rafiansyah (a*), Silvester Dian Handy Permana (a), Budi Arifitama (a)

a) Program Studi Teknik Informatika, Universitas Trilogi
Jl. TMP Kalibata No.1, Jakarta 12760, Indonesia
*mrafiansyah[at]trilogi.ac.id

Abstract

There are 500 mushrooms in the wild where 200 to 300 species can be consumed and some other species are not suitable for consumption even some are poisonous. But for the general public, it is certainly difficult to distinguish which mushrooms are worth eating and which are poisonous mushrooms. This study aims to identify the type of fungus using a backpropagation neural network as a model to determine the type of fungus and image processing as a method for retrieving data from digital images on each type of fungus. The backpropagation method will reduce the error rate in predicting the type of fungus assisted with Image Processing which will facilitate the input of mushroom characteristics. The number of parameters used as feature extraction is 9 which is divided into 6 color features and 3 texture features. The results of this study are 80 artificial neural network models with different parameters that use accuracy and MSE (Mean Square Error) as indicators of successful performance in each model in identifying the type of fungus. The highest accuracy value is 100% contained in 25 models of artificial neural networks which are divided into 4 models in the 20 max epoch model group, 5 models in the 30 max epoch group, 5 models in the 40 max epoch group and 11 models in the group with 50 max epochs.

Keywords: Artificial neural network, Backpropagation, Image Processing, Mushroom

Topic: Information Engineering