Forecasting of Vannamei Shrimp Production Based on Weather Factors Using Radial Basis Function Neural Network Approach (Case Study: Lamongan District)

Abstract

One of the economic activity sectors that can be influenced by uncertain weather conditions is aquaculture production. In Lamongan, aquaculture production especially vannamei shrimp is very dependent on the ideal weather conditions. Uncertainty of the weather conditions can cause irregular harvesting to the production of vannamei shrimp. These trend changes can have an impact on the activities of production supply chain, namely the fulfillment of the entry quota of vannamei shrimp production from agents or distributors to exporters of vannamei shrimp to meet market demand. This marketing result can increase the original income of Lamongan area. To find out the development of the trend required a forecasting process and appropriate classification based on past data using artificial neural networks. One structure of artificial neural networks that can be predicting and classifying is a radial basis function neural network (RBFNN). The structure of RBFNN is trained using K-means clustering and gradient descent method. We use average temperature, average humidity and rainfall each month starting from January 2013 until December 2017 as the actual datasets. From those datasets, the training datasets start from January 2013 until December 2016 and the remaining datasets are used as the testing datasets. Built in the Python program, the test results show that our forecasting and classification had an accuracy level with mean absolute percentage error (MAPE) 16.7%.