莊永坤REYNA RACHMA WATI2026-09-092026-09-092026-07-24https://handle.ncl.edu.tw/11296/ke8sq3https://203.71.86.71/handle/123456789/74556學位別：碩士 語文別：英文 指導教授：莊永坤 口試委員：廖凱威; 洪偉倫; 莊永坤 授權範圍：網際網路,開放日期為2026-07-30 電子論文連結：https://handle.ncl.edu.tw/11296/ke8sq3Fish is a highly nutritious and widely consumed food in Taiwan, however, its high moisture content and biological composition make it extremely perishable, requiring proper handling during storage and distribution process in order to avoid spoilage and quality deterioration. This work utilized near-infrared spectroscopy as a rapid, efficient, and non-destructive method to discriminate non-marinated, lime-marinated, and garlic-marinated Tilapia fillets based on the volatile nitrogen limit. Soft Independent modelling of Class Analogy (SIMCA) and Partial Least Squares Discriminant Analysis (PLS-DA) were employed to train the classification model using preprocessed spectra. PLS-DA gave superior predictive performance in predicting the samples, achieving 91.3% accuracy, 100% specificity, and 87.5% sensitivity, demonstrating that Near-Infrared Spectroscopy can be utilized to evaluate marinated Tilapia fish freshness and put a label based on the TVBN values between acceptable class for a value below 15 mg/100 g and non-acceptable class for a value above 15 mg/100 g which indicate a spoilaging process.系所名稱：食品安全學系碩士班thesis