Forecasting agricultural commodity pricing using neural network-based approach

Over the last decade, unprecedented spikes and drops in commodity prices have been a recurrent source of concern to both policymakers and the investors. This research paper focuses on effective prediction of commodities prices which will be a key contribution to the investment world and policy makers to devise strategies. Neural network and multiple regression models were built that would efficiently predict the price in advance for different forecasting ranges considering Kapas as the product. Historical prices along with various price influencing factors like inflation, rainfall, exchange rate and cottonseed oil cake price were given as input parameters to the feed forward multi-perceptron neural network and multiple regression models for forecasting prices. The results were compared using mean absolute percentage error (MAPE) as an accuracy measure. Artificial neural network models outperformed multiple linear regression models for medium-term and long-term data. The results indicate that for long-term predictions, neural network models have high predictive power.