-
Normalized GDP (monthly) https://fred.stlouisfed.org/
from 2000-01-01
to 2023-11-01 -
Libor Rates (daily) https://www.global-rates.com/en/
from 2001-01-02
to 2024-06-01 -
Current Account to GDP (quarterly) https://stats.oecd.org/Index.aspx?QueryId=67094#
from Q1-2000
to Q3-2024 -
Forex (daily) https://www.federalreserve.gov/
from 2000-01-03
to 2020-08-21from 2020-08-22
to 2024-06-15AUD EUR NZD GBP BRL CAD CNY DKK HKD INR JPY MYR MXN NOK ZAR SGD KRW LKR SEK CHF TWD THB VEB
Inflation has become a major and serious issue for countries around the world. The Real Effective Exchange Rate (REER) serves as a crucial indicator for evaluating a nation's international competitiveness and significantly impacts economic aspects such as exports, imports, investment, economic growth, and inflation. Fluctuations in REER's value present both opportunities and challenges, necessitating continuous monitoring and appropriate adjustments to maintain economic stability and development. As the world advances technologically and innovates, algorithms are being updated to comprehend the nature of unfolding phenomena. Machine learning (ML) algorithms are a contemporary phenomenon employed in various aspects of tasks. Real exchange rate data is considered a critical component of the business market, playing a pivotal role in understanding market trends. This study utilizes machine learning models, namely the Multilayer Perceptron (MLP), Convolution Neural Network (CNN), and classical time series models, including Autoregressive Integrated Moving Average (ARIMA) and Exponential Smoothing (ES), to model and forecast the real exchange rate (REER) dataset. The data under consideration spans from 2000 to 2024 and comprises 6126 observations. This study preprocesses and subsequently consolidates all this data. The study selects a model that meets the Key Performance Indicator (KPI) criteria. This chosen model is deemed the best candidate for forecasting the behavior of the real exchange rate dataset.
Keywords: REER, Forecasting, Machine learning, Multi-layer perceptron model, Exponential smoothing, CNN, ARIMA