Time Series Analysis of wheat production in Zimbabwe (1975-2023)

Author

Mutuva, Trouble.

Title

Time Series Analysis of wheat production in Zimbabwe (1975-2023)

Abstract

This study focused on time series analysis of wheat production in Zimbabwe from 1975 to 2023. In order to fulfil objectives of this study, FFNN and ARIMA models were developed based on annual wheat production data from 1975 to 2013(training set) and forecasts from 2014 to 2023 were deduced from the two models for comparisons with the actual data from 2014 to 2023(testing set). MAE and RMSE used as performance metrics. Results of this study show FFNN model as a best fit, since it comes out with lowest error values and it mimics the actual values of wheat production compared to ARIMA which shows a linear pattern over time and FFNN detected hidden patterns of wheat production data. Forecasts from the best selected FFNN model show an upward trend to be expected from 2024-2028, indicating that wheat production values are increasing, which is a good indicator to the agricultural sector. Based on the findings, this study recommends the use Artificial Neural Networks (ANN) for time series analysis and the government of Zimbabwe and policy makers to keep on enacting policies and support in agriculture to increase agricultural output and achieving the country’s Sustainable Development Goals (SDG) of being a middle-income country by 2030.

Date

June 2024

Publisher

BUSE

Keywords

Time series analysis

ARIMA models

FFNN models

Forecasting

Supervisor

Mr. B. Kusotera