Sea Animal Image Classification Using Machine Learning Algorithms for Accurate and Scalable Prediction

Abstract This study presents a comparative analysis of machine learning algorithms for classifying sea animal images using Orange Data Mining. The dataset, consisting of labeled images categorized into three classes—squid, statfish, and whale—was sourced from an open-access repository and processed using supervised learning workflows. Various models, including Neural Networks, Support Vector Machines (SVM), Random Forests, k-Nearest Neighbors (kNN), Logistic Regression, Naïve Bayes, Gradient Boosting, and AdaBoost, were evaluated using 10-fold cross-validation. Performance was assessed across multiple metrics: Area Under the Curve (AUC), Classification Accuracy (CA), F1-score, Precision, Recall, and LogLoss. The Neural Network model yielded the best overall performance with an AUC of 0.990 and a classification accuracy of 93.2%. SVM and Logistic Regression closely followed, outperforming other traditional and ensemble methods. Confusion matrix analysis further supported these findings, demonstrating low misclassification rates for Neural Networks. ROC curve evaluations for individual classes confirmed the robustness of top-performing models. The findings validate the effectiveness of low-code platforms like Orange in streamlining image classification pipelines for ecological and biological image datasets. This study provides valuable insights for researchers aiming to deploy interpretable and scalable machine learning solutions in marine biology and related domains.