Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

What’s the Price Tomorrow? - A Comparison of Artificial Neural Networks to Forecast EUR/SEK

Brinklert, Hannes LU (2025) NEKH01 20242
Department of Economics
Abstract
The future price of the FOREX rate is important for export-dependent countries and companies and the research regarding price prediction of FOREX is skewed towards the major currency pairs, leaving the smaller currency pairs to be unexplored. The study aimed to answer if the Multi-layer Perceptron (MLP) or the Long Short-Term Memory (LSTM) is the most suitable model for forecasting the rate of a small currency pair and if the models' predictions can be leveraged to create a simple profitable trading strategy. The MLP and the LSTM were developed and trained to predict the next day's EUR/SEK rate using past prices, economic data, and technical indicators. The two models were evaluated against a simple baseline predictor that predicts that... (More)
The future price of the FOREX rate is important for export-dependent countries and companies and the research regarding price prediction of FOREX is skewed towards the major currency pairs, leaving the smaller currency pairs to be unexplored. The study aimed to answer if the Multi-layer Perceptron (MLP) or the Long Short-Term Memory (LSTM) is the most suitable model for forecasting the rate of a small currency pair and if the models' predictions can be leveraged to create a simple profitable trading strategy. The MLP and the LSTM were developed and trained to predict the next day's EUR/SEK rate using past prices, economic data, and technical indicators. The two models were evaluated against a simple baseline predictor that predicts that tomorrow's price is today's price. Two trading strategies were created based on each model's predicted price movement and were evaluated against a simple trading strategy of only conducting long positions. The result shows that the simple baseline predictor was better at predicting the future FOREX rate than the two models and between the two, the MLP was slightly better. However, there is a clear relationship between the true price and the predicted price of the two machine learning models. The trading strategies display a different result where the MLP and LSTM-based strategies perform better than the long-only strategy. The MLP-based strategy was the best one regarding returns and Sharpe ratio. (Less)
Please use this url to cite or link to this publication:
author
Brinklert, Hannes LU
supervisor
organization
course
NEKH01 20242
year
type
M2 - Bachelor Degree
subject
keywords
Machine Learning, Artificial Neural Network, FOREX, Price Prediction, LSTM, MLP
language
English
id
9183720
date added to LUP
2025-05-08 09:15:36
date last changed
2025-05-08 09:15:36
@misc{9183720,
  abstract     = {{The future price of the FOREX rate is important for export-dependent countries and companies and the research regarding price prediction of FOREX is skewed towards the major currency pairs, leaving the smaller currency pairs to be unexplored. The study aimed to answer if the Multi-layer Perceptron (MLP) or the Long Short-Term Memory (LSTM) is the most suitable model for forecasting the rate of a small currency pair and if the models' predictions can be leveraged to create a simple profitable trading strategy. The MLP and the LSTM were developed and trained to predict the next day's EUR/SEK rate using past prices, economic data, and technical indicators. The two models were evaluated against a simple baseline predictor that predicts that tomorrow's price is today's price. Two trading strategies were created based on each model's predicted price movement and were evaluated against a simple trading strategy of only conducting long positions. The result shows that the simple baseline predictor was better at predicting the future FOREX rate than the two models and between the two, the MLP was slightly better. However, there is a clear relationship between the true price and the predicted price of the two machine learning models. The trading strategies display a different result where the MLP and LSTM-based strategies perform better than the long-only strategy. The MLP-based strategy was the best one regarding returns and Sharpe ratio.}},
  author       = {{Brinklert, Hannes}},
  language     = {{eng}},
  note         = {{Student Paper}},
  title        = {{What’s the Price Tomorrow? - A Comparison of Artificial Neural Networks to Forecast EUR/SEK}},
  year         = {{2025}},
}