Comprehensive Analysis of Exchange Rate Prediction: Influencing

Factors, Methodologies, and Future Prospects

Yunzhong Yu

School of Business, Macau University of Science and Technology, Macau, 999078, China

Keywords: Exchange Rate Prediction, Machine Learning Techniques, Deep Learning Techniques, Forecasting

Limitations.

Abstract: With the deepening of global economic integration and the increasing complexity of financial markets,

exchange rate forecasting has become increasingly important in economics and finance research. Exchange

rate fluctuations not only directly affect the international trade, but also profoundly impact investment

decisions and the formulation of monetary policies. Accurate exchange rate forecasts provide valuable

decision-making support for investors, policymakers, and enterprises, assisting them in managing market

fluctuations and economic uncertainty more effectively. This paper reviews the current research status in

exchange rate forecasting, focusing on the factors affecting exchange rates, forecasting methods based on

machine learning and deep learning, and the limitations of existing technologies. First, this paper

systematically introduces various economic factors that affect exchange rate fluctuations, including

macroeconomic indicators, market sentiment, political events, and international trade. Secondly, this paper

reviews existing exchange rate forecasting methods, including machine learning and deep learning techniques,

as well as their performance in forecasting accuracy and data processing capabilities, as well as their

advantages and limitations in practical applications. By deeply understanding the influencing factors and

optimizing the forecasting methods, the reliability of exchange rate forecasting can be significantly improved,

thereby providing more effective decision-making support for financial market participants and policymakers.

1 INTRODUCTION

An important factor in global economic trade is the

exchange rate, which is the value of one currency

relative to another. The price of a domestic currency

expressed in terms of a foreign currency is typically

expressed as the exchange rate, which represents the

relationship among the two currencies' exchange rates

(Cai, 2023). As an important bridge between

macroeconomics and microeconomics, the exchange

rate is often used as an important reference indicator

for evaluating the economic value of other countries.

The strength of economic fundamentals between

countries is the core factor in exchange rate changes

(Li, 2023) .

Fixed and variable exchange rates are the two

primary categories of exchange rates. The market's

supply and demand dynamics determine floating

exchange rates, which fluctuate in response to sudden

shifts in the overall state of the economy. This

https://orcid.org/0009-0008-2484-0318

exchange rate mechanism has high flexibility and can

be adjusted quickly to respond to dynamic changes in

the economy. In contrast, fixed exchange rates are set

by the government or central bank and maintained

within a stable range through market intervention,

thereby reducing the uncertainty caused by exchange

rate fluctuations. In addition, there are other types of

exchange rates, such as managed floating exchange

rates and currency union exchange rates. Managed

floating exchange rates combine the characteristics of

market determination and central bank intervention,

and maintain a certain degree of stability by adjusting

exchange rate fluctuations. Currency union exchange

rates refer to the use of a common currency by

multiple countries or regions to attain a high level of

economic integration and monetary policy

coordination. These different exchange rate Changes

in exchange rates also have a profound impact on all

aspects of a country's economy. First, exchange rates

directly affect the competitiveness of international

200

Yu, Y.

Comprehensive Analysis of Exchange Rate Prediction: Inﬂuencing Factors, Methodologies, and Future Prospects.

DOI: 10.5220/0013212800004568

In Proceedings of the 1st International Conference on E-commerce and Artiﬁcial Intelligence (ECAI 2024), pages 200-204

ISBN: 978-989-758-726-9

trade. A rise in the value of the currency may result in

higher export prices on the global market and a

potential decline in exports; on the other hand, a fall

in the value of the currency may make export items

more competitive and encourage export expansion.

Secondly, changes in exchange rates will also impact

the returns of cross-border investments. When

investors make international investments, exchange

rate risk becomes an important consideration, and

exchange rate fluctuations will directly affect their

investment returns. In addition, the execution of

national economic policy is significantly impacted by

the stability of the currency rate. Turbulence in the

financial market and economic uncertainty can result

from exchange rate volatility.

Under the general trend of economic globalization,

exchange rates are an important factor affecting

international trade and financial investment.

Scientific analysis of exchange rate fluctuations and

reasonable prediction of exchange rate fluctuations

have important theoretical research significance and

practical application value. Effective exchange rate

forecasting can not only help private investors adjust

their investment strategies and diversify risks

promptly, thereby optimizing investment returns but

also provide a key risk management decision-making

basis for multinational companies in international

transactions, so that investors and companies can

better cope with the challenges brought by exchange

rate fluctuations and reduce potential financial risks

(Wang et al., 2024).

This article summarizes the pertinent techniques

in the field of exchange rate forecasting, reviews the

findings of previous research, and assesses the

benefits and drawbacks of various machine learning

and deep learning algorithms in exchange rate

forecasting. These efforts are motivated by the

recognition of the critical role that exchange rate

forecasting plays in the global economy and the

significant influence it has on international trade,

investment choices, and financial market stability.

The factors influencing the exchange rate are first

introduced in this article, which also provides an

overview of the current methods for forecasting

exchange rates from the perspectives of deep learning

and machine learning. Finally, the limitations of these

methods are discussed, along with potential directions

for future research. By means of an extensive

examination from many angles, it seeks to offer a

thorough comprehension of the domain of exchange

2 EXCHANGE RATE FACTORS

Generally speaking, exchange rate fluctuations are

mainly affected by the size of foreign exchange

reserves (Li, 2021), inflation rate (Liu, 2020),

political situation (Miao, 2022), money supply (Wang

& Meng, 2019), interest rate (Yin, 2018), and global

economic conditions.

First and foremost, a nation's foreign exchange

reserves are a vital and necessary instrument for

controlling its economy and achieving both internal

and external balance. The economies of different

nations are entwined and readily impacted by the

economic activity of other nations in the setting of

global economic integration (Li, 2021). Therefore,

foreign exchange reserves have become an important

means of regulating the balance of payments. A larger

level of reserves can improve the economy's

resilience to risks and encourage the steady and

healthy growth of the economy by assisting in

managing global capital flows and economic shocks

as well as preserving exchange rate stability. Second,

the rate of inflation is a crucial metric for gauging the

extent of price growth over a given time frame, which

has an immediate impact on the real worth and buying

power of a nation's currency. More specifically, rising

inflation will cause domestic prices to rise (Liu, 2020),

making domestic goods more expensive in the

international market, thereby weakening the

competitiveness of export goods. At the same time,

inflation will also stimulate import demand because

foreign goods are relatively cheap, leading to a

widening trade deficit. In addition, inflation will

change the market's expectations of future economic

trends, weaken the confidence of investors and

trading partners in the local currency, and thus lead to

the depreciation of the local currency. The next factor

is the political situation (Miao, 2022). Political

instability or uncertainty will reduce investors'

confidence in a country and reduce their demand for

the country's currency, leading to capital outflow and

triggering exchange rate depreciation. On the

contrary, a stable political environment can enhance

investors' trust in the national economy, attract more

foreign investment, increase demand for the local

currency, and thus support exchange rate stability. A

stable political situation provides a more reliable

foundation for economic growth and helps maintain

the international credibility and value of the currency.

Then there is the money supply. The entire quantity

of money that is available for use in a nation's

economy throughout a specific time period is referred

to as the money supply. Changes in a nation's central

bank's monetary policy are typically closely linked to

Comprehensive Analysis of Exchange Rate Prediction: Inﬂuencing Factors, Methodologies, and Future Prospects

201

changes in the money supply. The quantity of money

in the market rises when a central bank implements

an accommodating monetary policy, such as cutting

interest rates or injecting significant sums of money.

The excess amount of money may cause the money

circulating in the market to exceed the actual demand

of the economy, leading to inflation, which in turn

drives up prices. The real purchasing power of local

currency decreases as inflation rises and the same

amount of money can buy fewer goods and services,

ultimately leading to a decline in the currency

exchange rate (Wang & Meng, 2019). Next is the

interest rate. Economist John Maynard Keynes

proposed the exchange rate parity theory, which

states that the difference in interest rates is equivalent

to the rate at which the exchange rate of two countries

changes. In the event of a disparity in interest rates

between two nations, money tends to go toward the

one with the higher rate in order to maximize profits.

Due to the profit-seeking nature of funds, it will lead

to a rise in the country's currency's demand with high

interest rates, thereby driving its currency to

appreciate. A nation with low interest rates

experiences a decline in demand for its currency at

the same time, which results in currency depreciation.

Consequently, the currency of the nation with lower

interest rates tends to weaken while the currency of

the nation with higher interest rates tends to

strengthen in the near run (Yin, 2018). Finally, the

global economic situation. Global economic growth

or recession will affect international trade and

investment flows. When the economy is strong, trade

and investment activities in various countries will

generally increase, and the demand for currency will

rise. Strong economic performance will generally

attract the interest of foreign investors, promote

capital inflows, and enhance the demand for a

country's currency, thereby increasing the exchange

rate. On the other hand, trade and investment activity

often decline during a recession, which lowers the

demand for money in a nation and lowers the value of

the currency.

3 EXCHANGE RATE

FORECASTING METHODS

3.1 Machine Learning Methods

Computer systems may learn from data and become

better versions of themselves thanks to a process

called machine learning. It is a part of artificial

intelligence that uses scientific methods such as

statistics, algorithms, and computers to identify

patterns in data to make predictions or decisions.

Specifically, machine learning methods can

automatically learn and identify patterns from large

amounts of data without the need for humans to write

complex rules. At the same time, it can continuously

improve its predictive capabilities and adapt to

changes in new data, that is, the machine learning

model can Historical data is used for forecasting and

handles various forecasting tasks such as

classification, regression, and clustering to provide

insights into future trends. Compared with artificial

learning, machine learning has significant advantages

in automation, processing of complex data, predictive

capabilities, and adaptability. It can improve the

performance and efficiency of the system by learning

from data and self-improvement, adapting to modern

data-driven application scenarios. Therefore, many

scholars use machine learning to study and analyze

exchange rate changes.

The theory of support vector machines was first

put forth by Vapnik (1995). Regression and

classification issues can be handled by support vector

machines. Compared with neural network theory, this

model is based on statistical principles, emphasizes

minimizing structural risks, and has stronger

generalization ability. This method can better handle

the prediction problem of exchange rate out of sample,

is suitable for solving nonlinear problems, and is also

suitable for small sample data sets (Zhao, 2021). A

novel approach based on the blockchain framework

and XGBoost was put out by Shahbazi and Byun

(2022) to increase the system's security and

transparency. Filters and coefficient weights are used

in the construction of its prediction process, and the

XGBoost algorithm exhibits the highest accuracy in

the test. Fu et al. (2019) proposed two optimized

support vector machine (SVR) models based on phase

space reconstruction of historical exchange rate data

to predict four typical RMB exchange rates. The SVR

parameters are adjusted by balancing global and local

optimization. The model has good results in

horizontal prediction accuracy, directional prediction

accuracy, and statistical accuracy.

3.2 Deep Learning Methods

Deep learning focuses on using neural networks

(especially deep neural networks) for data processing

and pattern recognition. Through brain-like

simulation, it automatically pulls features and makes

predictions or classifications from massive amounts

of data. Typically, deep learning models have several

layers of processing units, with each layer pulling out

ECAI 2024 - International Conference on E-commerce and Artiﬁcial Intelligence

202

more advanced characteristics from the one above it.

Its advantages are the ability to process unstructured

data, efficient pattern recognition, strong

generalization and adaptability, and simplified

modeling. An increasing number of academics have

turned to deep learning techniques in order to

estimate exchange rates in recent years. Although

these studies have adopted a variety of different

methods, deep learning is still the most widely used

in exchange rate forecasting.

Liu et al. (2024) suggested combining the LASSO

and LSTM models to identify the variables

influencing exchange rate pricing and provide logical

justifications. A novel LASSO-BiLSTM integrated

learning technique for variables influencing exchange

rate prices was put out by them. Compared with

traditional models such as ELM, KELM, LSTM, and

SVR, LASSO-BiLSTM performs better than other

models and shows good results in time series

prediction. Windsor and Cao (2022) developed an

algorithm model based on the multimodal fusion MF-

LSTM model to predict the USD/CNY exchange rate.

Paying attention to market indicators and investor

sentiment builds a deep coupling model. This method

shows that it is effective in integrating multimodal

fusion algorithms into financial time series prediction.

Liu et al. (2023) proposed a hybrid model CNN-

STLSTM-AM to predict the exchange rate's closing

price, in which CNN extracts data features, STLSTM

improves the prediction accuracy and combines the

attention mechanism (AM) to optimize the feature

weights. Empirical evaluations demonstrate that the

CNN-STLSTM-AM model surpasses other

techniques, including as CNN, SVR, LSTM, CNN-

LSTM, GRU-LSTM, and CNN-LSTM-AM, in terms

of prediction accuracy. Wang et al. (2021) suggested

a CNN-TLSTM model to estimate the closing price

of the USD/RMB exchange rate on the following

trading day in order to quickly grasp the changeable

exchange rate information. In order to prevent

overfitting, it enhances the input gate structure of the

LSTM and extracts features using CNN. The

experimental findings demonstrate the effectiveness

of the CNN-TLSTM model in forecasting the

USD/RMB exchange rate's closing price for the

following trading day. Combining the views of

investors and investment organizations, Chen et al.

(2021) creatively presented a dual-objective

optimization measurement model for portfolio

exchange prediction analysis. The deep learning

model offers more precise forecasts than the

conventional exchange rate prediction method, and

the NSGA-II-based model further improves portfolio

selection. It has better prediction accuracy than

traditional models and can effectively help investors

make good investment decisions.

4 EXISTING LIMITATIONS AND

FUTURE CHALLENGES

Existing studies have conducted a relatively complete

and comprehensive analysis of exchange rate

forecasting, but there are still certain limitations. First,

there are many factors that affect exchange rates.

Medium- and long-term exchange rates are more

affected by fundamental factors, including fiscal

policy, government intervention measures, inflation,

international trade, etc., and there is a high degree of

uncertainty in the medium and long term. Many

factors that affect exchange rates, such as

emergencies and political changes, lack high-

frequency data or reliable real-time information and

are difficult to measure effectively, which makes it

more difficult to establish an accurate forecasting

model (Yin, 2018). Second, exchange rate changes

may have complex nonlinear patterns, and data may

fluctuate greatly. This volatility makes the model

learning less effective and will be affected by more

unstable data, thereby affecting the accuracy of the

forecast. In addition, there may be time delays in the

release of economic data and market indicators, and

these delays may affect the timeliness of exchange

rate forecasts. Delayed data may not reflect the latest

market changes in a timely manner, making the

model's forecast results unable to accurately reflect

the current economic situation.

The following adjustments may be made in the

future in light of the aforementioned restrictions. First,

in-depth research on fundamental factors such as

fiscal policy, government intervention measures,

inflation, and international trade, and develop a

forecasting model that can comprehensively consider

the impact of these factors. Minimize the impact of

potential uncertainties on medium- and long-term

exchange rate forecasts. In addition, develop more

advanced data collection tools to capture and

integrate more high-frequency, real-time economic

and market data, including information on

emergencies and political changes, to ensure that the

model can obtain and process the latest economic and

market data in a timely manner. Then promotes the

openness, transparency and standardization of

economic data, reduces the impact of economic data

delays on exchange rate forecasts, and better conducts

data analysis and model development. Finally, more

models can be used for integration, and more complex

Comprehensive Analysis of Exchange Rate Prediction: Inﬂuencing Factors, Methodologies, and Future Prospects

203

deep learning and multi-model integrated learning

methods can be used to reduce forecast bias and

further increase the stability and accuracy of forecasts.

5 CONCLUSIONS

This paper deeply explores the key elements of

exchange rate forecasting, including its basic

concepts, influencing factors, types, two different

exchange rate forecasting methods and their current

limitations, and looks forward to future development

trends. Exchange rate forecasting is not only of great

significance to international financial markets,

strategic decisions of multinational companies and

global economic stability, but also affects the

decision-making process of investors and

policymakers. The substance of this article begins

with an explanation of the fundamental idea of an

exchange rate, which is the rate at which one currency

is exchanged for another. Exchange rates come in two

flavors: fixed exchange rates and fluctuating

exchange rates. Then, by reviewing the factors

affecting exchange rates, this paper reveals how

factors such as foreign exchange reserves, inflation

rate, political situation, money supply and interest

rate interact and affect exchange rate fluctuations.

Additionally, this article highlights the

accomplishments and benefits of previous research in

the subject of exchange rate forecasting and discusses

the use and development of deep learning and

machine learning techniques in this area. The

shortcomings of the current exchange rate forecasting

techniques are also highlighted in this research,

including high uncertainty in exchange rate

influencing factors, large fluctuations in data, delays

in data and lack of timeliness. Therefore, although

advanced forecasting technologies have made

important progress in improving forecasting accuracy,

these limitations still need to be overcome. Finally,

this paper points out that in the future, the researchers

should execute in-depth research on economic

fundamentals and develop forecasting models that

can comprehensively consider the impact of these

factors. Secondly, the researchers should improve the

effectiveness of data collection and promote the

openness and transparency of economic data to

ensure that the model can obtain and process the latest

economic and market data in a timely manner, to

continuously optimize the performance of the model

and provide more reliable forecasting results in a

complex and changing economic environment,

thereby further enhancing its application value in

global economic decision-making.

REFERENCES

Cai, H., 2023. Application research on foreign exchange

rate prediction based on deep learning and transfer

learning (Master's thesis, Beijing University of Civil

Engineering and Architecture).

Chen, J., Zhao, C., Liu, K., Liang, J., Wu, H., Xu, S., 2021.

Exchange Rate Forecasting Based on Deep Learning

and NSGA-II Models. Computational Intelligence and

Neuroscience, 2021(1), 2993870.

Fu, S., Li, Y., Sun, S., Li, H., 2019. Evolutionary support

vector machine for RMB exchange rate forecasting.

Physica A: Statistical Mechanics and its Applications,

521, 692-704.

Li, J., 2023. Research on exchange rate prediction based on

CNN-BiLSTM and attention mechanism (Master's

thesis, Fujian Normal University).

Liu, G., 2020. Factors influencing RMB exchange rate

fluctuations and empirical analysis. Times Finance, 28,

68-70, 79.

Li, H., 2021. Analysis of factors influencing RMB

exchange rate fluctuations. National Circulation

Economy, 29, 139-141.

Liu, P., Wang, Z., Liu, D., Wang, J., Wang, T., 2023. A

CNN-STLSTM-AM model for forecasting usd/rmb

exchange rate. Journal of Engineering Research, 11(2),

100079.

Liu, S., et al., 2024. A new LASSO-BiLSTM-based

ensemble learning approach for exchange rate

forecasting. Engineering Applications of Artificial

Intelligence.

Miao, Y., 2022. Research on exchange rate prediction based

on machine learning and ensemble models (Master's

thesis, Shandong University).

Shahbazi, Z., Byun, Y., 2022. Knowledge discovery on

cryptocurrency exchange rate prediction using machine

learning pipelines. Sensors, 22(5).

Wang, G., Chen, H., Ma, J., Wang, J., 2024. Research on

multi-step exchange rate prediction based on multi-

scale 1D-CNN and attention mechanism. Systems

Engineering Theory and Practice, 06, 1934-1949.

Wang, J., Wang, X., Li, J., Wang, H., 2021. A prediction

model of CNN-TLSTM for USD/CNY exchange rate

prediction. Ieee Access, 9, 73346-73354.

Wang, S., & Meng, Q., 2019. Analysis of factors

influencing RMB real exchange rate fluctuations.

Economic Research Guide, 23, 138-141.

Windsor, E., Cao, W., 2022. Improving exchange rate

forecasting via a new deep multimodal fusion model.

Applied Intelligence, 52(14), 16701-16717.

Yin, C., 2018. Study on factors affecting RMB exchange

rate trends (Master's thesis, Shanghai Jiao Tong

University).

Zhao, B., 2021. Combination research of machine learning

in exchange rate prediction (Master's thesis, Zhejiang

University).

ECAI 2024 - International Conference on E-commerce and Artiﬁcial Intelligence

204