Klasifikasi Citra Aktivitas Sport, Casual Dan Leisure Berbasis Machine Learning

SUSANTI, Emilia (2024) Klasifikasi Citra Aktivitas Sport, Casual Dan Leisure Berbasis Machine Learning. Undergraduate thesis, Universitas Katolik Widya Mandira Kupang.

	Text ABSTRAK.pdf Download (1MB)
	Text BAB I.pdf Download (417kB)
	Text BAB II.pdf Restricted to Repository staff only Download (587kB)
	Text BAB III.pdf Restricted to Repository staff only Download (688kB)
	Text BAB IV.pdf Restricted to Repository staff only Download (455kB)
	Text BAB V.pdf Restricted to Repository staff only Download (404kB)
	Text BAB VI.pdf Download (331kB)
	Text DAFTAR PUSTAKA DAN SURAT KETERANGAN HASIL CEK PLAGIASI.pdf Download (443kB)

Abstract

In the ever-evolving digital era, documentation has an important role in recording and sharing information related to daily activities, including sports, casual, and leisure activities. Therefore, the classification of documentation becomes important to help identify between the three activities. This study aims to classify images that depict sport, casual, and leisure activities using the KNN method. 600 image data was taken from two sources,namely the internet and direct capture using a mobile phone camera, each with 300 images with a data division of 200 images for each activity category. The selection of the value of K=3 is considered appropriate because it provides an optimal balance. The validation results showed that K=3 with a fold of 20 provided good performance with an accuracy value of 85.6%, F1-score 85.4%, precision 86.5% and recall 85.6% also showed good performance. The confusion matrix analysis revealed the model's ability to distinguish activity categories, although there were some classification errors, especially between casual and sports classes. This study highlights the importance of comprehensive metric evaluation in evaluating the performance of classification models. In conclusion, the use of the KNN method to classify the image of sport, casual, and leisure activities results in good performance, with significant improvements when using the number of folds and the selection of the right K value.

Actions (login required)

View Item