Pacific saury fishing forecasting by using random forest

作者：Baba Shinya^*; Matsuishi Takashi

来源：Nippon Suisan Gakkaishi, 2015, 81(1): 2-9.

DOI：10.2331/suisan.81.2

In Japan, Pacific saury fishing forecasts are indispensable information for fisheries. In this study, the abundance of Pacific saury in fishing grounds was predicted by using random forest and the forecast was evaluated by the Monte Carlo re-sampling method. A Pacific saury abundance index with three categories was used for the response variables. A total of 186 explanatory variables from 22 indices of marine environments from 1972 to 2011 were used. After variable selection by random forest, only four kinds of variable were chosen as explanatory variables. The hitting ratio became approximately 62% and exceeded the accuracy of present posted forecasts. By using random forest, the explanatory variables can be compressed and accurate forecasts can be made by using many explanatory variables.

• 出版日期2015-1