CAJ | 학술논문

为了能更好预测西北太平洋柔鱼的资源量，选择合适的预测方法及开发相应的预测系统颇为重要。利用相关性分析，筛选出在产卵区显著影响西北太平洋柔鱼资源量的关键网格点，并采用这些网格点的海表温度、产卵区适宜温度所占面积的比例和单位努力捕获量等数据组织样本，然后利用线性回归、BP神经网络、RBF 神经网络和支持向量机等预测方法进行实验。结果表明：在西北太平洋柔鱼中长期预测中， BP神经网络要优于其他方法。以相关性分析和BP神经网络为基础建立的西北太平洋柔鱼资源量预测系统是有效可行的。
위료능경호예측서북태평양유어적자원량，선택합괄적예측방법급개발상응적예측계통파위중요。이용상관성분석，사선출재산란구현저영향서북태평양유어자원량적관건망격점，병채용저사망격점적해표온도、산란구괄의온도소점면적적비례화단위노력포획량등수거조직양본，연후이용선성회귀、BP신경망락、RBF 신경망락화지지향량궤등예측방법진행실험。결과표명：재서북태평양유어중장기예측중， BP신경망락요우우기타방법。이상관성분석화BP신경망락위기출건립적서북태평양유어자원량예측계통시유효가행적。
In order to better predict resource of squid in the Northwestern Pacific, it is very important to develop a prediction system based on an appropriate forecasting method. The correlation analysis was used to filter out key mesh points that significantly affect resource of squid in spawning areas. Data samples were organized by using sea-surface temperature (SST) at these mesh points, the ratio between the area with suitable temperature and the whole spawning areas, and catching amount per unit of effort (CPUE). Prediction experiments have been conducted by using linear regression, BP network, RBF network and support vector machine. Results showed that BP neural network is much better than other methods during medium and long term prediction of squid in the Northwestern Pacific. The resource forecasting system built based on correlation analysis and BP neural network is effective and feasible.