COMPARATIVE STUDY OF ARIMA AND ARTIFICIAL NEURAL NETWORK MODELS IN FORCASTING FOOD PRICES

Abstract

Prices of food items are known to be a good indicator of food security and knowledge of their future value cannot be overemphasized. Box-Jenkins ARIMA model is a frequently used approach in forecasting time series. However, recent researches suggest that Artificial Neural Network (ANN) model can be a good alternative to the conventional use of ARIMA model. This study seeks to examine and compare the forecast efficiency of ARIMA and ANN models for data on food prices in Nigeria. The best ARIMA and ANN models was selected for each of the prices of the different food items out of all the suggested models. Results obtained revealed that ANN models had the minimum Root Mean Square Error (RMSE) and Mean Absolute Percentage Error (MAPE) when compared with ARIMA for all of the food items, hence the conclusion that ANN models produce more efficient forecasts than the Box-Jenkins ARIMA model for predicting food prices in Nigeria. The study also recommended more investment in local agricultural production and enlightenment of the public on the importance of diet diversification to dampen the price effect on the Nigerian populace.