BUDIMAN, Naufal and Adi, Kusworo and Wibowo, Adi (2025) OPTIMASI CONVOLUTIONAL NEURAL NETWORK DENGAN PENDEKATAN DIFFERENTIAL EVOLUTION DALAM KLASIFIKASI KEMATANGAN BUAH KELAPA SAWIT. Masters thesis, UNIVERSITAS DIPONEGORO.

	Text Cover aja.pdf Download (208kB)
	Text Cover lengkap.pdf Restricted to Repository staff only Download (1MB)
	Text Bab 1.pdf Download (300kB)
	Text Bab 2.pdf Download (683kB)
	Text Bab 3.pdf Restricted to Repository staff only Download (654kB)
	Text Bab 4.pdf Restricted to Repository staff only Download (1MB)
	Text Bab 5.pdf Restricted to Repository staff only Download (282kB)
	Text Daftar Pustaka.pdf Download (289kB)

Abstract

Penentuan tingkat kematangan buah kelapa sawit secara akurat merupakan aspek penting dalam menjaga kualitas hasil panen dan efisiensi proses produksi minyak sawit. Penelitian ini mengusulkan model klasifikasi citra buah kelapa sawit berdasarkan tingkat kematangannya (mentah, matang, busuk) menggunakan arsitektur Convolutional Neural Network (CNN) yang dioptimasi dengan algoritma Differential Evolution (DE). Penggunaan DE bertujuan untuk meningkatkan performa CNN melalui pemilihan optimal hyperparameter dan konfigurasi arsitektur model. Dataset terdiri dari 302 citra buah kelapa sawit yang dikumpulkan secara langsung dari lapangan dan telah melalui proses augmentasi untuk mengatasi keterbatasan jumlah data. Penelitian ini mengembangkan model DE-CNN dan membandingkannya dengan baseline CNN serta tiga model transfer learning populer: EfficientNetB3, ResNet50, dan DenseNet121. Evaluasi kinerja dilakukan menggunakan metrik akurasi, presisi, recall, dan F1-score. Hasil menunjukkan bahwa model DE-CNN memberikan performa yang sangat baik dengan nilai akurasi 94,62%, precision 95,37%, recall 94,62%, dan F1-score 94,65%. Selain itu, sistem klasifikasi ini juga diimplementasikan dalam bentuk aplikasi berbasis web, memungkinkan pengguna untuk mengunggah citra buah dan langsung memperoleh hasil klasifikasi. Penelitian ini menunjukkan bahwa integrasi CNN dan DE secara efektif meningkatkan akurasi klasifikasi kematangan buah kelapa sawit.
Kata Kunci : Convolutional Neural Network, Differential Evolution, Klasifikasi Citra, Kematangan Kelapa Sawit, Optimasi Hyperparameter, Deep Learning

Accurate determination of oil palm fruit ripeness is a crucial aspect in maintaining harvest quality and ensuring the efficiency of palm oil production processes. This study proposes an image classification model for oil palm fruit based on ripeness levels (unripe, ripe, rotten) using a Convolutional Neural Network (CNN) architecture optimized with the Differential Evolution (DE) algorithm. The use of DE aims to enhance CNN performance by selecting optimal hyperparameters and architectural configurations. The dataset consists of 302 oil palm fruit images collected directly from the field and augmented to address the limitation in data quantity. This research develops a DE-CNN model and compares its performance with a baseline CNN and three popular transfer learning models: EfficientNetB3, ResNet50, and DenseNet121. Performance evaluation is conducted using accuracy, precision, recall, and F1-score metrics. The results demonstrate that the DE-CNN model achieves strong performance, with an accuracy of 94.62%, precision of 95.37%, recall of 94.62%, and an F1-score of 94.65%. Furthermore, the classification system is implemented as a web-based application, allowing users to upload fruit images and receive immediate classification results. This study indicates that the integration of CNN and DE effectively improves the accuracy of oil palm fruit ripeness classification.
Keywords: Convolutional Neural Network, Differential Evolution, Image Classification, Oil Palm Ripeness, Hyperparameter Optimization, Deep Learning

Actions (login required)

View Item