A Comparative Analysis Of Time Series And Neural Networks Models In Forecasting Zwl/Usd Exchange Rates.

Author

Ishmael Mudenda

Title

A Comparative Analysis Of Time Series And Neural Networks Models In Forecasting Zwl/Usd Exchange Rates.

Abstract

This research is a comparative analysis of FFNN and ARIMA model in forecasting exchange rate volatility in Zimbabwe. The study applied FFNN [1(5,5)1] and ARIMA (0,1,2) models in forecasting ZWL/USD exchange rates volatility using performance evaluation techniques such as RMSE, Symmetric MAPE, AIC, and BIC. The major objective was to compare the performance of these models in predicting future exchange rates. The research used data of weekly exchange rates extracted from the Reserve Bank of Zimbabwe’s website to come up with models. Data used was for the period February 2022 to December 2023. Forecasting was based on In-Sample and Out-Of-Sample predictive horizon. The research findings selected FFNN model since it had the lowest Symmetric MAPE, RMSE, AIC, BIC and RMS values. The FFNN model predicted values were adjacent to the actual data. Based on the findings, the study indorses the use FFNN model for forecasting exchange rates since it can predict large volumes of data in a given data set.

Date

June 2024

Publisher

BUSE

Keywords

Time Series Forecasting

Exchange Rates

ARIMA,

Neural Network

Supervisor

Ms J Pagana