Skip to main content

LUP Student Papers

LUND UNIVERSITY LIBRARIES

ARMAX-modelling with Google Trends data for prediction of changes in volume in cryptocurrencies

Berner, Willem LU and Lennér, Caspar (2025) In Bachelor's Theses in Mathematical Sciences FMSL01 20251
Mathematical Statistics
Abstract (Swedish)
This thesis examines the use of ARMAX models for predicting cryptocurrency trading volumes, with Google Trends data serving as an exogenous variable. The spec- ulative nature of cryptocurrencies suggests that market sentiment, reflected in search trends, could influence trading activity. To test this hypothesis, the predictive performance of ARMAX models is compared to ARMA models across six cryptocurrencies, using metrics such as RMSE and rise-and-fall accuracy. The results indicate that ARMAX models do not consistently outperform ARMA models. The confidence intervals of the measured parameters indicating performance does not let us draw any conclusions. The ARMA model seems to cover most of the model dynamics, but in some of the cases... (More)
This thesis examines the use of ARMAX models for predicting cryptocurrency trading volumes, with Google Trends data serving as an exogenous variable. The spec- ulative nature of cryptocurrencies suggests that market sentiment, reflected in search trends, could influence trading activity. To test this hypothesis, the predictive performance of ARMAX models is compared to ARMA models across six cryptocurrencies, using metrics such as RMSE and rise-and-fall accuracy. The results indicate that ARMAX models do not consistently outperform ARMA models. The confidence intervals of the measured parameters indicating performance does not let us draw any conclusions. The ARMA model seems to cover most of the model dynamics, but in some of the cases the exogenous inputs supply significance.
This study contributes to the field of financial modeling by evaluating the role of alternative data in time series forecasting and providing a nuanced perspective on its applicability in cryptocurrency markets. (Less)
Please use this url to cite or link to this publication:
author
Berner, Willem LU and Lennér, Caspar
supervisor
organization
course
FMSL01 20251
year
type
M2 - Bachelor Degree
subject
publication/series
Bachelor's Theses in Mathematical Sciences
report number
LUTFMS-4016-2025
ISSN
1654-6229
other publication id
2025:K1
language
English
id
9182066
date added to LUP
2025-01-20 13:50:16
date last changed
2025-02-11 14:25:54
@misc{9182066,
  abstract     = {{This thesis examines the use of ARMAX models for predicting cryptocurrency trading volumes, with Google Trends data serving as an exogenous variable. The spec- ulative nature of cryptocurrencies suggests that market sentiment, reflected in search trends, could influence trading activity. To test this hypothesis, the predictive performance of ARMAX models is compared to ARMA models across six cryptocurrencies, using metrics such as RMSE and rise-and-fall accuracy. The results indicate that ARMAX models do not consistently outperform ARMA models. The confidence intervals of the measured parameters indicating performance does not let us draw any conclusions. The ARMA model seems to cover most of the model dynamics, but in some of the cases the exogenous inputs supply significance.
This study contributes to the field of financial modeling by evaluating the role of alternative data in time series forecasting and providing a nuanced perspective on its applicability in cryptocurrency markets.}},
  author       = {{Berner, Willem and Lennér, Caspar}},
  issn         = {{1654-6229}},
  language     = {{eng}},
  note         = {{Student Paper}},
  series       = {{Bachelor's Theses in Mathematical Sciences}},
  title        = {{ARMAX-modelling with Google Trends data for prediction of changes in volume in cryptocurrencies}},
  year         = {{2025}},
}