Time Series Modelling of Yearly Cassava Production in Nigeria: A Comparative Study

Cassava production is an important agricultural activity in Nigeria, as it contributes to the GDP of the polity. Appropriate prediction of cassava production in the nation Nigeria is fundamental to the development of a long-term plan to sustain agricultural productivity and promote food security. This study investigated in detail statistical characteristics of yearly cassava production in Nigeria over the period 1961 to 2022 with a view to choosing a befitting model for the data. The data set was divided into training set and test set. By virtue of ADF test, the training set was found to be nonstationary. The Zivot-Andrew test revealed the presence of a structural break in the data. The break date was found to be 1990. Holt’s linear model with multiplicative errors, ARIMA (1,1,2) model and SETAR (2,2,1) model were fitted to the training set following their automatic selection using ets, auto.arima and Selectsetar functions in the forecast and tsDyn packages in R. The out of sample comparison of the three models based on their associated root mean squared errors (RMSEs) and mean absolute percentage errors (MAPEs) provided the evidence of the SETAR mode having the smallest RMSE and MAPE values. Hence, SETAR (2,2,1) model is the best for forecasting annual cassava production in Nigeria among the three models.