Classification of Dragon Fruit Stem Diseases Using Convolutional Neural Network

Abstract

A holticulture plant known as dragon fruit (pitaya) is a fruit that has many benefits and is widely cultivated by farmers in several areas of Banyuwangi. In dragon fruit plants there are various kinds of diseases that attack including red spot, stem rot, black rot, scab, and mosaic. Farmers still recognize diseases on dragon fruit stems manually so that sometimes there are errors in disease recognition. In this research, a system was developed to identify the types of diseases on dragon fruit stems. This system was built by proposing the Convolutional neural network method with the proposed architecture using the Python programming language with the Tensorflow, Keras, and Scikit-Learn libraries. The proposed system is tested using k-fold cross validation with tunning parameters fold = 5 and epoch = 5. The training results show that the highest accuracy performance value is 85.06% with the data used as test data as many as 191 images producing 147 correct data and 44 data wrong, while the average overall accuracy score was 76.43%.

Teknik Random Forest untuk Meningkatan Akurasi Data Tidak Seimbang

“…SMOTE dapat meningkatkan kinerja algoritma machine learning di kelas minoritas dan mengurangi resiko overfitting di kelas mayoritas [8]. Overfitting adalah kondisi dimana hasil akurasi tinggi pada masa training, tetapi rendah pada saat testing [9], [10].…”