Implementasi Artificial Neural Network (ANN) Backpropagation Untuk Klasifikasi Jenis Penyakit Pada Daun Tanaman Tomat

Abstract

Tomato is a horticultural plant that can be easily found in Indonesia. Tomatoes are popular for public consumption because their nutritional values have excellent content for health, besides tomatoes are widely cultivated because the price is quite stable. However, there are obstacles in tomato plants cultivation for farmers because tomatoes are vulnerability to diseases, so that farmers have difficulty distinguishing types of diseases that look similar. Therefore, this research was conducted to help farmers in identifying various diseases in tomato leaves. This research uses Artificial Neural Network (ANN) method with backpropagation algorithm to classify the types of diseases in tomato leaves. The data used for testing were 50 image data of bacterial spot disease, 50 image data of yellow leaf curl disease, and 50 image data of healthy leaves. Tomato leaves types of diseases classification with the best result that utilized backpropagation with Cross-validation 4 folds  by following the rules with batch size of 100, hidden layers defined as , the learning rate for weight updates is 0.3, with validation threshold is used to terminate validation testing is 20, and 500 as the number of epoch to train through, have been resulted in an accuracy of 78% which required time 319.77 seconds to process data. Which was then assisted by Confusion Matrix as a tool to measure the performance of the classification results that have been carried out and precision results of 0.78 so as to obtain 117 data of true positive.

Keywords: Tomato, Leaf imagery classification, Artificial Neural Network (ANN), Backpropagation.