Feature Selection for Explaining Yellowfin Tuna Catch per Unit Effort Using Least Absolute Shrinkage and Selection Operator Regression

Abstract

To accurately identify the key features influencing the fisheries distribution of Pacific yellowfin tuna, this study analyzed data from 43 longline fishing vessels operated from 2008 to 2019. These vessels operated in the Pacific Ocean region (0° to 30° S; 110° E to 170° W), with a specific focus on 25 features of yellowfin tuna derived from marine environment data. For this purpose, this study opted for the Lasso regression analysis method to select features to predict Pacific yellowfin tuna fishing grounds, exploring the relationship between the catch per unit effort (CPUE) of yellowfin tuna and multiple features. This study reveals that latitude and water temperature at various depths, particularly the sea surface temperature of the preceding and subsequent months and the temperature at depths between 300 and 450 m, are the most significant features influencing CPUE. Additionally, chlorophyll concentration and large-scale climate indices (ONI and NPGIO) also have a notable impact on the distribution of CPUE for yellowfin tuna. Lasso regression effectively identifies features that are significantly correlated with the CPUE of yellowfin tuna, thereby demonstrating superior fit and predictive accuracy in comparison with other models. It provides a suitable methodological approach for selecting fishing ground features of yellowfin tuna in the Pacific Ocean.