Comparative Analysis of ARIMA and Artificial Neural Network Techniques for Forecasting Non-Stationary Agricultural Output Time Series

With the vast popularity of deep learning models in engineering and mathematical fields, Artificial Neural Networks (ANN) have recently attracted significant research applications agriculture, economics, informatics finance. In this paper, we use a method to capture predict unknown complex nonlinear characteristics agricultural output based on autoregressive artificial neural network, using Nigeria as case study. Using proposed model, shocks is analyzed modeled data obtained for period forty years (1980-2019), compared with analyses from integrated moving average model (ARIMA). This result because it justifies superiority hybrid ANN over traditional Box-Jenkins methodology forecasting non-stationary time series. The empirical results show that achieves an improved accuracy ARIMA method. It further various types networks would be useful solving relevant tasks problems widely defined global production.