Analisis performa metode Hybrid Convolutional Neural Network dan K-Nearest Neighbors pada klasifikasi kesegaran daging sapi

Mubarok, Haikal Mufid (2026) Analisis performa metode Hybrid Convolutional Neural Network dan K-Nearest Neighbors pada klasifikasi kesegaran daging sapi. Sarjana thesis, UIN Sunan Gunung Djati Bandung.

Preview	Text COVER_Haikal Mufid Mubarok_1217050059.pdf Download (153kB) | Preview
Preview	Text Abstrak_Haikal Mufid Mubarok_1217050059.pdf Download (180kB) | Preview
Preview	Text pernyataan karya sendiri.pdf Download (67kB) | Preview
Preview	Text DAFTAR ISI_Haikal Mufid Mubarok_1217050059.pdf Download (216kB) | Preview
Preview	Text BAB I_Haikal Mufid Mubarok_1217050059.pdf Download (261kB) | Preview
	Text BAB II_Haikal Mufid Mubarok_1217050059.pdf Restricted to Registered users only Download (729kB) | Request a copy
	Text BAB III_Haikal Mufid Mubarok_1217050059.pdf Restricted to Registered users only Download (288kB) | Request a copy
	Text BAB IV_Haikal Mufid Mubarok_1217050059.pdf Restricted to Registered users only Download (643kB) | Request a copy
	Text BAB V_Haikal Mufid Mubarok_1217050059.pdf Restricted to Registered users only Download (179kB) | Request a copy
	Text DAFTAR PUSTAKA_Haikal Mufid Mubarok_1217050059.pdf Restricted to Registered users only Download (174kB) | Request a copy

Abstract

Meat freshness assessment is a crucial aspect of ensuring food safety and quality. Convolutional Neural Networks (CNN), particularly the Xception architecture, have proven effective in automatic visual feature extraction. However, the integration of deep learning feature extraction with classical classification algorithms remains to be explored for optimal performance. This study aims to develop and analyze a hybrid method combining CNN Xception as a feature extractor and K-Nearest Neighbors (KNN) as a classifier for beef freshness classification into three categories: fresh, half-fresh, and spoiled. The research follows the CRISP-DM methodology using a primary dataset of 806 .jpg images. Testing was conducted with variations of the K parameter (K=3, K=5, and K=7) using Euclidean Distance. Evaluation results show that the hybrid CNN Xception-KNN model achieved the best performance at K=7 with an accuracy of 94.44%, precision of 0.94, recall of 0.94, and F1-score of 0.94. These findings demonstrate that utilizing Xception as a feature extractor combined with KNN provides high and consistent accuracy in image-based beef quality detection.

Actions (login required)

View Item