Preliminary study for the evaluation of the hematological blood parameters of seabream with machine learning classification methods

Abstract

Fish production becomes more and more critical in the world. The most important parameters in production are health, safety, and economy. As a requirement of the technological age, computer-based applications are utilized in providing all these, for example, artificial neural networks, machine learning, deep learning, …etc. In this study, Random Forest (RF), k-Nearest Neighborhood (k-NN), Naive Bayes (NB), Logistic Regression (LR), Decision Tree (DT) and Support Vector Machines (SVM), were used by Python 3.6 programming language for the evaluating the effects of mannanoligosaccharides with hematological blood parameters. Two diets (0%, and 2‰) prepared for the fish. At the end of 12 months trial period, there were no significant changes in the hematological blood parameters without hematocrit (Hct). Hct decreased significantly in fish fed with 2‰ mannanoligosaccharides. RF, LR, DT, and SVM have the same highest accuracy level with 80 %. However, k-NN and NB have the same lowest accuracy level with 60 %. According to these results, RF, LR, DT, and SVM models can be used as a classification tool between groups. Thus, it might be that the sensitivity level of correctly chosen machine learning applications is higher than statistical analysis.