<![CDATA[Purpose: Applying the Convolutional Neural Network method of VGG16 architecture and determining the right hyperparameter combination and knowing the performance of the CNN algorithm accuracy level in classifying Mount Merapi images.Design/Method/Approach: The Mount Merapi dataset was split with a ratio of 80:10:10 for the division of training, validation, and testing data. Furthermore, 360 x 360 pixels image resizing and 224 x 224 cropping were carried out with the aim of adjusting to the input in the VGG16 model, and min-max normalization was carried out. Next, modeling is done using CNN VGG16 architecture with a hyperparameter testing approach of epoch 10, 20, 30, 40 and 50 with Adam, RMSprop, and SGD optimizers. Then the evaluation is done using confusion matrixFindings/result: The test results have been carried out with hyperparemeter epoch 10, 20, 30, 40 and 50 with Adam optimizer, RMSprop, and SGD on CNN architecture VGG16. This study obtained an optimal model using epoch 50 with Adam's optimizer which resulted in an accuracy value on test data of 99.15%.Originality/value/state of the art: Applying a combination of testing optimization algorithms and hyperparameters can result in accuracy that can reach the optimal point.]]>
Copyrights © 2024
