Application of Artificial Neural Networks to forecast Litopenaeus vannamei and Penaeus monodon harvests in Indramayu Regency, Indonesia

• Published in: IOP conference series. Earth and environmental science Vol. 521; no. 1; pp. 12018 - 12025
• Main Authors: Pamungkas, A, Zulkarnain, R, Adiyana, K, Waryanto, Nugroho, H, Saragih, A S
• Format: Journal Article
• Language: English
• Published: Bristol IOP Publishing 01.06.2020
• Subjects: Algorithms; Aquaculture; Artificial intelligence; Artificial Neural Networks; Back propagation; Back propagation networks; Correlation coefficient; Correlation coefficients; Crustaceans; Economic forecasting; Environmental impact; Forecasting; Government agencies; Litopenaeus vannamei; Machine learning; Neural networks; Penaeus monodon; Root-mean-square errors; Indonesia
• ISSN: 1755-1307; 1755-1315
• DOI: 10.1088/1755-1315/521/1/012018

Besides minimizing environmental impact, one of the goals of ecological intensification for aquaculture is production. Production forecasting is needed to make policies in planning, especially in terms of meeting consumer demand. This paper introduces a method to forecast the total shrimp production for Litopenaeus vannamei and Penaeus Monodon in Indramayu Regency using artificial neural networks. In this case, we used backpropagation neural networks (BPNN). BPNN is a supervised learning algorithm and usually used by perceptron with many layers to change the weights associated with the neurons in the hidden layers. During the training process, the network calculated the output that will be generated based on the given input patterns. The network assigned and adjusted the weights of the input and also the hidden layer to obtain a network with good performance. Networks with small error values close to zero indicate good performance. The criteria used to test the performance of the artificial neural networks method are the root mean squared error (RMSE), the mean absolute percentage error (MAPE), and the correlation coefficient (r). Production data obtained from the relevant government agencies were used to train the algorithms as a part of an artificial intelligence process. This artificial intelligence forecasted the shrimp's harvest. Forecasting performance is indicated by the accuracy of the prediction process data compared to the real data. The best result for L. vannamei forecasting was obtained in the trainGD with MSE 0.0174 and MAPE 19.28%. The best results P. monodon forecasting were obtained in the TrainRP with MSE 0.0200 and MAPE 22.99%.