Forecasting of Export Prices of Indian Shrimp: An ANN Approach

Abstract

Both internal and external externalities create impediments in price determination. This is more so when a product like the Indian-farmed shrimp is entirely an export commodity. Risk and uncertainty loom large over Indian shrimp farmers at both the farm and end-consumer levels. It becomes necessary to minimize the volatility in forecasted prices to help shrimp farmers stabilize their production strategies reasonably well. Export prices of Indian Pacific white and black tiger shrimp for different count sizes were analysed for developing forecasting models for two major international markets, viz., the United States and Japan. Overall, 16 data sets were compiled for developing appropriate price forecasting models. Models were developed to forecast weekly prices of shrimp exports from India and import prices from the United States and Japan. Out of several artificial neural network (ANN) models attempted, the best ANN models were selected based on the Akaike information criterion. The ANN model captured the fibrillations and fluctuations and gave a close prediction with actual values. Since the forecasted values were accurate and in sync with real data, the values can be expected to give a correct representation of the behaviour of the markets.