Analisis Efektivitas MobileNetV2 dalam Klasifikasi Motif Batik dengan Menggunakan Hyperparameter Tuning dan Model Optimizer

Rafli, Muhammad (2026) Analisis Efektivitas MobileNetV2 dalam Klasifikasi Motif Batik dengan Menggunakan Hyperparameter Tuning dan Model Optimizer. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) Muhammad Rafli_Laporan Skripsi-1-16.pdf Download (785kB)
	Text (Bab 1) Muhammad Rafli_Laporan Skripsi-17-26.pdf Download (272kB)
	Text (Bab 2) Muhammad Rafli_Laporan Skripsi-27-62.pdf Restricted to Repository staff only until 23 January 2029. Download (1MB)
	Text (Bab 3) Muhammad Rafli_Laporan Skripsi-63-70.pdf Restricted to Repository staff only until 23 January 2029. Download (841kB)
	Text (Bab 4) Muhammad Rafli_Laporan Skripsi-71-114.pdf Restricted to Repository staff only until 23 January 2029. Download (1MB)
	Text (Bab 5) Muhammad Rafli_Laporan Skripsi-115-116.pdf Download (206kB)
	Text (Daftar Pustaka) Muhammad Rafli_Laporan Skripsi-117-122.pdf Download (183kB)
	Text (Lampiran) Muhammad Rafli_Laporan Skripsi-123-126.pdf Restricted to Repository staff only until 23 January 2029. Download (307kB)

Abstract

Digital image-based batik motif classification is a challenging task in computer vision due to the high visual similarity between motifs and complex variations in shape and color. Convolutional Neural Networks (CNNs) have been widely applied for image classification tasks; however, conventional CNN architectures often suffer from high computational complexity and require substantial hardware resources. Therefore, this study aims to analyze the effectiveness of MobileNetV2 as a lightweight CNN architecture for batik motif classification and to investigate the impact of hyperparameter tuning and optimizer selection on model performance. The dataset used in this study consists of 2,462 batik motif images categorized into 13 classes. The data were split into training, validation, and test sets, with the test set fixed from the beginning of the experiment. The proposed framework incorporates data augmentation, stratified sampling with five iterations, and a two-phase training scheme consisting of transfer learning and fine-tuning. Hyperparameter tuning was performed using four search methods: grid search, random search, bayesian optimization, and particle swarm optimization (PSO), combined with four optimizers: Adam, SGD, RMSprop, and Adagrad. The optimized hyperparameters include learning rate, batch size, and dropout rate. Experimental results indicate that the combination of bayesian optimization and the Adam optimizer achieves the best performance, with a mean validation accuracy of approximately 91–92 percent and low validation loss. Evaluation on the test set yields an overall accuracy of 92.05 percent, with high precision, recall, and f1-score across all classes and a well-distributed error pattern in the confusion matrix. Furthermore, the analysis of hyperparameter search time reveals a trade-off between computational cost and performance improvement. Overall, this study demonstrates that MobileNetV2, when optimized through appropriate hyperparameter tuning and optimizer selection, provides accurate, stable, and computationally efficient performance for batik motif classification.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Prasetya, Dwi Arman	NIDN0005128001	arman.prasetya.sada@upnjatim.ac.id
Thesis advisor	Hindrayani, Kartika Maulida	NIDN0009099205	kartika.maulida.ds@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA76.6 Computer Programming

Divisions:

Faculty of Computer Science > Departemen of Data Science

Depositing User:

Muhammad Rafli

Date Deposited:

28 Jan 2026 04:41

Last Modified:

28 Jan 2026 04:41

URI:

https://repository.upnjatim.ac.id/id/eprint/49025