Machine Learning - Based Prediction of Genotoxicity in Peripheral Erythrocytes of Fish (Labeo catla): A Comparative Analysis of Algorithms

Kousik Seal, Soumendra Nath Talapatra

Abstract: This study utilizes machine learning ML algorithms to predict the accuracy of a genotoxicity dataset, focusing on nuclear abnormalities in peripheral erythrocytes of Labeo catla. Eight ML algorithms, including Logistic Regression, K - nearest neighbour, Lazy. KStar, DecisionStump, Hoeffding Tree, RandomForest, and RandomTree, were tested using the WEKA tool. Among these, RandomForest demonstrated the highest predictive accuracy with an area under the ROC curve of 91%. These results indicate that ML algorithms, particularly RandomForest, provide an effective approach for predicting genotoxicity in fish species.

Keywords: Edible fish, Genotoxicity dataset, Machine learning, MN & NA parameters, Labeo catla