인공지능을 이용한 주요 수산물 가격 예측 모형 비교에 관한 연구

Comparative Study on Price Forecasting Models of Major Fisheries Products Using Artificial Intelligence

The purpose of this study is to establish a model for predicting the fluctuations of the frozen wholesale market prices of five major consumption fish species such as mackerel, hairtail, pollock, squid, and yellow corvina using five AI machine learning algorithms such as Decision Tree, Random Forest, Gradient-Boost, XG-Boost, and SVM, and to compare the predictive powers with each other using various forecasting indicators. The case of best prediction power was the prediction of the price of hairtail using a random forest, where the accuracy was 0.923, even more showing 100% precision, especially in the case of price decline. Among the five algorithms, the highest predictive power was SVM, with an average accuracy at 0.683, while the lowest one was XG-Boost, with an average accuracy at 0.614. When comparing the predictive powers of the algorithm for each individual fish species, Gradient-Boost and SVM were the best for mackerel, decision tree and random forest for hairtail, and random forest and XG-Boost for pollack. In addition, the decision trees was found to be the algorithms with the highest predictive power for squid, just like SVM was for yellow corvina.