Classification of Chicken Meat Freshness Using Color and Texture Feature Fusion with Bayesian Optimization Cross Validation on LightGBM

Saka, Agatha Diani Putri Saka (2026) Classification of Chicken Meat Freshness Using Color and Texture Feature Fusion with Bayesian Optimization Cross Validation on LightGBM. Undergraduate thesis, UPN Veteran Jawa Timur.

	Text (Cover) 22081010094.-cover.pdf Download (887kB)
	Text (Bab 1) 22081010094.-bab1.pdf Download (228kB)
	Text (Bab 2) 22081010094.-bab2.pdf Restricted to Repository staff only until 15 June 2028. Download (464kB)
	Text (Bab 3) 22081010094.-bab3.pdf Restricted to Repository staff only until 15 June 2028. Download (843kB)
	Text (Bab 4) 22081010094.-bab4.pdf Restricted to Repository staff only until 15 June 2028. Download (1MB)
	Text (Bab 5) 22081010094.-bab5.pdf Download (95kB)
	Text (Daftar Pustaka) 22081010094.-daftarpustaka.pdf Download (183kB)
	Text (Lampiran) 22081010094.-lampiran.pdf Restricted to Repository staff only Download (2MB)

Abstract

This study evaluates the effect of Bayesian Optimization on the performance of the LightGBM algorithm for classifying chicken meat freshness from digital images. The research process included dataset preparation, image resizing to 256 × 256 pixels, data augmentation, Region of Interest (ROI) extraction, and data splitting using ratios of 80:20, 70:30, and 60:40. Feature extraction was performed using three feature combinations, namely HSV-GLCM, HSV-LBP, and HSV-GLCM LBP. These features were then used as inputs for both the standard LightGBM model and the optimized model. Chicken meat samples were categorized into three classes: fresh, less fresh, and rotten. Model effectiveness was measured using accuracy, precision, recall, F1-score, and AUC. The experimental results indicate that the optimized LightGBM model generally produced better classification results than the standard model. The highest performance was obtained from the HSV-GLCM-LBP feature combination with an 80:20 data split, achieving an accuracy of 90.15%, precision of 90.18%, recall of 90.15%, F1-score of 90.09%, and AUC of 0.9707. The findings demonstrate that combining color, texture, and local texture information provides a more representative feature set for chicken meat freshness classification. Under the same feature configuration, the standard LightGBM model achieved an accuracy of 86.01%, indicating a performance improvement of 4.14% after optimization. The best performing model was also integrated into a web based application capable of automatically predicting chicken meat freshness from uploaded images. Overall, Bayesian Optimization effectively improved LightGBM performance and supported the development of an automated freshness classification system.

Item Type:

Thesis (Undergraduate)

Contributors:

Contribution	Contributors	NIDN/NIDK	Email
Thesis advisor	Muttaqin, Faisal	NIDN0030058602	faisalmuttaqin.if@upnjatim.ac.id
Thesis advisor	Anggraeny, Fetty Tri	NIDN0711028201	fettyanggraeny.if@upnjatim.ac.id

Subjects:

Q Science > QA Mathematics > QA76.87 Neural computers

Divisions:

Faculty of Computer Science > Departemen of Informatics

Depositing User:

Agatha Diani Putri Saka

Date Deposited:

19 Jun 2026 07:03

Last Modified:

19 Jun 2026 07:44

URI:

https://repository.upnjatim.ac.id/id/eprint/54006