Publicação
Predictive Modelling of the Bitcoin Price: A Comprehensive Analysis of Time Series Models - The usage of time-based models in predicting the price of Bitcoin in both the short and long term future
| datacite.subject.fos | Ciências Naturais::Ciências da Computação e da Informação | pt_PT |
| dc.contributor.advisor | Pinheiro, Flávio Luís Portas | |
| dc.contributor.author | Costa, Ana Rita Pires da | |
| dc.date.accessioned | 2024-03-08T11:43:58Z | |
| dc.date.available | 2024-03-08T11:43:58Z | |
| dc.date.issued | 2024-01-31 | |
| dc.description | Project Work presented as the partial requirement for obtaining a Master's degree in Information Management, specialization in Knowledge Management and Business Intelligence | pt_PT |
| dc.description.abstract | The extraordinary volatility of Bitcoin is attributed to a multitude of external factors that are difficult to identify and monitor for predictive modeling. The impact of the Bitcoin halving introduces still another level of complexity, implying possible stability in the far future, subject to discernible seasonality indicators. This thesis aims to identify what variables influence the value of bitcoin and what kind of models better forecast its price. Following the CRISP-DM methodology, data was collected from online sources, analyzed, and treated accordingly as means to be fed to different types of models: Autoregressive Integrated Moving Average (ARIMA), Vector Autoregressive (VAR), Recurrent Neural Network (RNN) with Long Short-Term Memory (LSTM) Layers and Prophet. The variables that were found to have the greatest potential for inclusion in prediction models were Bitcoin volume, the VT index, and currencies like the Israeli New Shekel and Euro, as well as silver and copper. Although Prophet showed potential, it was clear that it had limitations, especially when it came to predicting long-term outcomes with variables other than Bitcoin. Furthermore, in these models, regressors appeared to be a promising technique, but could only be used to anticipate known futures. While the RNN model with LSTM layers seemed reliable for short-term forecasts, it was not practical for making large-scale, long-term investment decisions. These results highlight how difficult it is currently to develop predictive models that accurately predict Bitcoin values, particularly for large-scale, long-term decision-making in the investment market. | pt_PT |
| dc.identifier.tid | 203543785 | pt_PT |
| dc.identifier.uri | http://hdl.handle.net/10362/164639 | |
| dc.language.iso | eng | pt_PT |
| dc.rights.uri | http://creativecommons.org/licenses/by/4.0/ | pt_PT |
| dc.subject | Time series forecasting | pt_PT |
| dc.subject | Price Prediction | pt_PT |
| dc.subject | ARIMA | pt_PT |
| dc.subject | VAR | pt_PT |
| dc.subject | RNN | pt_PT |
| dc.subject | LSTM Layers | pt_PT |
| dc.subject | Prophet | pt_PT |
| dc.subject | SDG 8 - Decent work and economic growth | pt_PT |
| dc.subject | SDG 9 - Industry, innovation and infrastructure | pt_PT |
| dc.title | Predictive Modelling of the Bitcoin Price: A Comprehensive Analysis of Time Series Models - The usage of time-based models in predicting the price of Bitcoin in both the short and long term future | pt_PT |
| dc.type | master thesis | |
| dspace.entity.type | Publication | |
| rcaap.rights | openAccess | pt_PT |
| rcaap.type | masterThesis | pt_PT |
| thesis.degree.name | Mestrado em Gestão de Informação, especialização em Gestão do Conhecimento e Inteligência de Negócio (Business Intelligence) | pt_PT |