Application of radial basis Function Neural Network for fishery forecasting

Abstract

In this paper, Radial Basis Function Neural Network is presented for fishery forecasting which uses Southwest Atlantic Illex argentines as its testing ground. The model begins with obtaining the network parameters to train the model using training data set and eventually achieving the forecasting results using test data set. The centre for basis function are selected from training set, weights of basis function for optimizing the fit of network is determined by orthogonal least square (OLS) method. In this paper, altogether six environmental factors are used which are months, longitude and latitude, sea surface temperature (SST), Sea surface Height (SSH) and chlorophyll for predicting the Total Habitat Index. The predicted values obtained are in terms of Total habitat index, which is calculated from two different indices such as Job number index and Average daily production index. The statistical model, Multiple Linear regressions is also implemented for fishery forecast. The results obtained from the RBFNN model were compared with Multiple Linear regressions in terms of accuracy criterions MSE, RAE ad PE. It is shown that the intelligent model has high predictive ability and better goodness of fit with respect to statistical models.