NIFS announced that in September 2024 it published its research outcome in Fishes, an international, scientific, open access journal published monthly online: Method to rapidly and correctly identify length at sexual maturity of fish by utilizing machine learning - a subset of artificial intelligence (AI). 

In the research, NIFS estimated the maturity stages and length of small yellow croaker – the major commercial fish species in South Korea - utilizing various machine learning algorithms (Decision Tree, Random Forest, LightGBM, XGB, SVM, etc.) based on ecological big data on the species that had been accumulated over a long time. 

The length at maturity provides an important basis for identifying ecological characteristics of fish and critical and scientific evidence for setting the body length that is prohibited from catching as one of the policies for fisheries resources management. 

In the exisiting method, it took much time and efforts as it used visual observation or tissue identification. 

Whereas this new method let AI identify the stages and estimate the maturity length autonomously based on its learning of the massive data including body length, weight, maturity, catch timing, etc. 

By doing so, the AI method not only improved the accuracy* of identification but also largely reduced the time** for analysis. It also eliminated the subjective criteria, making the identification continue to have consistency and high creidibility. 

   * (Existing method: maturity length 142) confidence interval 11.8~15.6cm → (AI method: maturity length 152) confidence interval 14.1~15.9cm

   ☞ That is, the range obtained by subtracting the lower limit from upper limit became narrowed. (3.8cm → 1.8cm)

   ** With the steps for identification skipped, the total time for measurement was reduced by around 30% (3 hours → 2 hours per fish species) 

The research outcome was the first case in Korea to estimate the maturity length of small yellow croaker by utilizing AI. 

The method is considered to have a high value as it is applicable to various fish species and able to provide scientific data important in fisheries resources management in Korea.