Authors:
Modupe Stella Omotayo-Tomo, Deebom Zorle Dum, Lekara-Bayo Ifeoma Better, Agbeye Isaac Oluwashola, Ahmad Albloush, Dang Van My
Addresses:
Department of Mathematical Sciences, Olusegun Agagu University of Science and Technology, Okitipupa, Ondo State, Nigeria. Department of Statistics, Rivers State University, Port Harcourt, Rivers State, Nigeria. Department of Mathematics, Rivers State University, Port Harcourt, Rivers State, Nigeria. Department of Business, Amman Arab University, Amman, Amman Governorate, Jordan. Department of Marketing, University of Finance - Marketing, Duc Nhuan Ward, Ho Chi Minh City, Vietnam.
Abstract:
This study examines linear, non-linear, and combined predictive models to forecast food prices in Nigeria using monthly time-series data from the World Food Programme spanning 15 January, 2002, to 15 July, 2026. Descriptive statistical analysis indicates a right-skewed distribution ("skewness"=1.46) with an average value of 3,644.99, a standard deviation of 3,226.45, and a Jarque–Bera test result that is highly significant (182.28,p=2.62×10^(-40)), rejecting the assumption of normality. CUSUM test results (3.0232, p=2.30×10^(-8)) and Bai–Perron analysis of structural fractures suggest continuous regime shifts among sub-periods (2002-2026). ADF (-3.7686, p = 0.003$) and PP (-3.638, p = 0.005) unit root assessments confirm first-difference stationarity (I (1)). Based on the lowest information criteria (AIC = 5073.054, BIC = 5084.105, HQIC = 5077.480) and diagnostics for white-noise residuals, ARIMA (1,1,1) was the best linear model. In the out-of-sample assessment, the standalone ARIMA model had significant forecast errors (RMSE = 1793.56, MAE = 843.73, MAPE = 41.15%), while the standalone LSTM had even larger errors (MAPE = 67.74%). They showed moderate gains, with the Hybrid ARIMA–ANN model achieving RMSE = 1425.53 and MAPE = 31.68%, and the Hybrid ARIMA–LSTM model achieving RMSE = 1808.26 and MAPE = 41.35%. The Artificial Neural Network (ANN) has the best prediction accuracy and error metrics (RMSE = 1318.93, MAE = 594.87, MAPE = 26.02%).
Keywords: Artificial Neural Network (ANN); Structural Breaks; World Food Programme (WFP); Jarque–Bera Test; Nonlinear Dynamics; Information Criteria; Forecast Errors.
Received on: 15/05/2025, Revised on: 20/07/2025, Accepted on: 01/09/2025, Published on: 05/06/2026
DOI: 10.64091/ATITP.2026.000335
AVE Trends in Intelligent Technoprise Letters, 2026 Vol. 3 No. 2 , Pages: 101-115