This paper addresses the problem of fish escapes in trout farming within the context of climate risk management for the aquaculture industry, with the objective of cost minimization for Kaizen projects. To address this problem, two different approaches are proposed: (I) a robust optimization-based approach and (II) a random forest-based approach. The robust approach aims to minimize cost by focusing on the worst-case scenario, while the random forest approach predicts the cost associated with the number of escaped fish by considering the levels (high and low) of features affecting the response variable. A case study from an enterprise in the Black Sea region of Turkey, a significant exporter of seafood, is presented to illustrate the problem and its external risk causes. The effectiveness of both approaches is evaluated through multiple instances. The results reveal that the random forest-based approach should be preferred when 60–70% of the features have low levels. Conversely, the robust approach should be preferred when most of the features have high levels to reduce the cost incurred by group-based Kaizen ap-plications. The findings highlight the potential of these methodologies to enhance sustainability in the aquaculture industry. Future research directions include extending the problem to cover different stages of the aquaculture supply chain and adopting other machine learning algorithms to improve system performance.