Advanced Predictive Modeling for Sweet Potato Production Estimation in Mozambique: A Comparative Analysis of ARIMA and LSTM Models for Supporting Food Security
Filipe Mahaluca*, Faizal Carsane, Rabeca Cuna and Alfeu Vilanculos
ABSTRACT
This study evaluated the application of two predictive models, ARIMA and LSTM neural networks, to estimate sweet potato production in Mozambique, aiming to provide accurate forecasts to support agricultural planning and food security. Sweet potato is a crucial crop in Mozambique, significantly contributing to the food subsistence of much of the rural population. Given the vulnerability of agricultural production to climate conditions and other external factors, accurate food production forecasting is vital for mitigating food insecurity in the country. The methodology included the use of the ARIMA model to capture linear patterns and historical trends from 1961 to 2009, with validation conducted between 2010 and 2020. The LSTM model, on the other hand, was trained with data from 1961 to 2013 and validated between 2014 and 2022. This model was chosen for its ability to identify complex and nonlinear patterns, offering greater accuracy in agricultural contexts with high variability. Forecasts for the period from 2023 to 2030 were generated using both models, focusing on providing insights that could support strategic decision-making in the agricultural sector. The results demonstrated that the LSTM model outperformed ARIMA in terms of accuracy, presenting a significantly lower Mean Absolute Percentage Error (MAPE), indicating greater effectiveness in predicting sweet potato production. Projections for the 2023–2030 period indicate stable production, with slight annual variations but no significant growth. This reflects resilient agriculture, but also highlights the need for strategic interventions to increase production and meet growing food demand. In conclusion, the LSTM model proved to be a more effective tool for forecasting agricultural production in scenarios of high uncertainty and variability, such as in Mozambique. The stable forecasts provided by this study underscore the importance of improving agricultural practices and investing in infrastructure and technologies to ensure that sweet potato production contributes sustainably to the country’s food and nutritional security.
