Modeling export price of tea in Kenya: Comparison of artificial neural network and seasonal autoregressive integrated moving average

Abstract

Agriculture sector is a key driver of economic growth in Kenya. It remains the main source of livelihood for the majority of the Kenyan people. Tea, coffee, and horticulture are the main agricultural exports in Kenya. Export price of these commodities fluctuates mainly due to law of demand and supply. Other reasons include; quality of goods and inflation effect on the dollar or other hard currencies. Further, farmers and their cooperative societies are affected by the local foreign exchange. The government and other stake holders require prior information on price trends for ease of planning. Thus it is important to forecast export prices of these commodities. The purpose of this study is to compare the forecasting performance of artificial neural network (ANN) model and a SARIMA model in export price of tea in Kenya. Secondary data was obtained from Kenya National Bureau of Statistics (KNBS). A total of 185 monthly export prices were obtained. A three layer feed-forward artificial neural network was trained using 70% of the data. The ANN model obtained was used to predict export prices for the remaining 30% of the data. SARIMA model was also used to predict export prices for the same duration. Forecasting performance was evaluated using Root mean squared errors (RMSE), mean absolute error (MAE) and mean absolute percentage error (MAPE). ANN demonstrated a superior performance over SARIMA model. The authors' ANN has high performance compared to SARIMA and can accurately predict export price of tea. Keywords: Artificial Neural Network (ANN), Seasonal Autoregressive Integrated Moving Average (SARIMA), Kenya National Bureau of Statistics (KNBS)