Klasifikasi Biji Kering Kakao Forastero Terfermentasi dari Berbagai Daerah di Indonesia Menggunakan Metode Convolutional Neural Network (CNN) dan Support Vector Machine (SVM)

Kristianingsih, Wahyu and Prof. Dr. Ir. Bambang Dwi Argo,, DEA. and La Choviya Hawa,, STP, MP, Ph.D. and Misnawi, Dr. Ir. (2023) Klasifikasi Biji Kering Kakao Forastero Terfermentasi dari Berbagai Daerah di Indonesia Menggunakan Metode Convolutional Neural Network (CNN) dan Support Vector Machine (SVM). Magister thesis, Universitas Brawijaya.

Abstract

Kakao Forastero sebagai varietas yang dikenal mudah dalam perawatanya, banyak dibudidayakan di berbagai wilayah Indonesia diantaranya yaitu daerah provinsi Aceh, Kalimantan Barat, Sumatera Barat, Sulawesi Barat, Bali, Banten dan Daerah Istimewa Yogyakarta. Berkaitan dengan hal tersebut, konsistensi yang tinggi terhadap pemenuhan standar mutu kakao perlu dilakukan agar tercapai keseragaman produk, serta sebagai upaya penanganan terhadap kecurangan melalui tindak pencampuran biji unggul dengan kualitas yang lebih rendah. Sehingga proses klasifikasi kedalam beberapa kelas yang sesuai berdasarkan ciri fisik tanpa proses destruktif dilakukan dengan memanfaatkan Support Vector Machine (SVM) dan Convolutional Neural Network (CNN) sebagai metode pembangunan model klasifikasi dengan arsitektur pemodelan yang digunakan diantaranya SquezeeNet, AlexNet, GoogleNet dan ResNet, serta untuk SVM dengan memanfaatkan algoritma kernel linier, kernel polynomial dan kernel RBF. Hasil menunjukkan bahwa pemodelan CNN dan SVM telah berhasil dibangun melalui dua tahap penelitian utama yaitu proses training-validation dan proses testing yang diterapkan pada input berupa citra bitmap. Secara umum, pemodelan CNN mampu menghasilkan performansi terbaik dengan akurasi training-validation sebesar 99,93% pada arsitektur SqueezeNet. Sedangkan untuk arsitektur SVM hanya mampu menghasilkan akurasi sebesar 93,49% pada arsitektur kernel Linier. Berdasarkan nilai testing yang diperoleh, arsitektur ResNet CNN menjadi pemodelan terbaik dalam penelitian ini dengan akurasi sebesar 99,99% lebih baik dibandingkan akurasi pada pemodelan SVM yang menghasilkan 56% melalui penggunaan kernel Linier.

English Abstract

Cocoa Forastero is a variety that easy to care because this type of cocoa plants is more resistant to pest, has high productivity and is widely cultivated in various parts of Indonesia, including the provinces of Aceh, West Kalimantan, West Sumatra, West Sulawesi, Bali, Banten and the Daerah Istimewa Yogyakarta. Therefore to fulfil the domestic and especially for the export market, it is necessary to have high consistency. Also an effort to deal with the act of mixing superior seeds with lower quality, it can be done through the practice of sorting or classifying products into several classes based on physical characteristics without destructive processes. So that Support Vector Machine (SVM) and Convolutional Neural Network (CNN) were chosen as the model building method with several modeling architectures to be used including SquezeeNet, AlexNet, GoogleNet and ResNet. As well for SVM, a modeling architecture has been applied with three types of kernels, namely linear kernels, polynomial kernels and RBF kernels. The results show that the CNN and SVM modeling have been successfully built through two main research stages, the training-validation process and the testing process applied to the bitmap images as input data. In general, CNN modeling is able to produce the best performance with a training-validation accuracy of 99.93% on the SqueezeNet architecture. At the same time the SVM architecture is only able to produce an accuracy of 93.49% on the Linear kernel architecture. Based on the testing values, the ResNet CNN architecture is the best modeling in this study with an accuracy of 99.99%. This figure is better than the accuracy obtained by the SVM modeling, because it only produces 56% in Linear kernels.

Item Type: Thesis (Magister)
Identification Number: 0423100009
Uncontrolled Keywords: Kakao Forastero, Klasifikasi, Deep Learning, Machine Learning,Cacao Forastero, Clasification, Deep Learning, Machine Learning, Convolutional Neural Network, Support Vector Machine Convolutional Neural Network, Support Vector Machine
Subjects: 300 Social sciences > 338 Production > 338.1 Agriculture > 338.16 Production efficiency
Divisions: S2/S3 > Magister Keteknikan Pertanian, Fakultas Teknologi Pertanian
Depositing User: Sugeng Moelyono
Date Deposited: 15 Aug 2023 03:10
Last Modified: 15 Aug 2023 03:10
URI: http://repository.ub.ac.id/id/eprint/202452
[thumbnail of DALAM MASA EMBARGO] Text (DALAM MASA EMBARGO)
Wahyu Kristianingsih.pdf
Restricted to Registered users only until 31 December 2025.
Download (2MB)

Actions (login required)

View Item View Item