Hyperparameter Optimized Rapid Prediction of Sea Bass Shelf Life with Machine Learning

Abstract

The article focuses on the importance of sea bass, which is preferred by consumers in Turkey and worldwide. However, seafood can deteriorate rapidly under unfavorable conditions during storage due to their nutrient content, water content, and weakness in connective tissues. Temperature changes, inappropriate processing methods during transportation, and temperature changes during storage in markets are reported to cause losses in seafood quality. The deterioration of seafood, especially in seafood stored under inappropriate conditions because of temperature, causes changes contrary to consumer preferences because of the rapid growth of microorganisms, especially odor changes in seafood. This study examines the models related to the discipline of predictive microbiology, which are stated to provide an accurate shelf life prediction of the rate of microbiological spoilage and emphasize the importance of mathematical predictions of these models for seafood. Furthermore, the paper observes that machine learning algorithms such as Random Forest, Decision Tree, k-Nearest Neighbors, AdaBoost, Gradient Tree Boosting, Random Forest, Decision Tree, k-Nearest Neighbors, AdaBoost, and Gradient Tree Boosting have been used to predict the shelf life of seafood products. Finally, how to augment the limited data in a laboratory study to evaluate the shelf life of sea bass stored at different temperatures, how to prove the consistency of the augmented data with the original data, and how to optimize successful machine learning methods for robust problem-solving processes between different engineering fields are explained in detail. The results show that the optimized Extra Tree algorithm is the most successful for Pseudomonas quantity estimation with an R² metric value of 0.9940 and TVC quantity estimation with an R² metric value of 0.9910, while the other algorithms are less successful than this algorithm. These results show that machine learning methods can be a rapid, powerful, and effective tool for shelf life prediction of sea bass. Additionally, it should be emphasized that the number of input parameters (temperature, number of the bacteria) are of utmost significant for augmentation of the data for development and application of the machine learning algorithms.