Publicação
A CASE STUDY ON THE EFFECT OF NEWS ON CRUDE OIL PRICE
| Resumo: | Crude oil price volatility has an impact on the global economy and oil-dependent industries and is influenced by supply and demand, geopolitical tensions, and the global economy. Every day, a massive amount of textual information flows in the form of news articles, which humans use to forecast future trends. News articles can have a significant impact on the price of crude oil because they contain information about recent events, trends, and advancements in the industry. The purpose of this work is to investigate how news articles may affect crude oil prices, using the concept of topic modeling and its potential for handling data. Using the webscraping method, the data for the study comes from a large dataset of news articles about the crude oil industry. These news articles were published between January 1 and December 31, 2022, and come from four different sources. The data was compiled using the source Exchange Rates UK to demonstrate how the price of crude oil fluctuated during this period. After the cleaning process was completed, the dataset contained a total of 1532 news articles. The Latent Dirichlet Allocation (LDA) technique is suggested for extracting relevant keywords from news articles and then using the findings as input features to forecast the crude oil price. The forecasting methods employed in the study were the Ridge model, the Random Forest and XGBoost techniques, and the time series method ARIMAX. The outcomes of the experiment indicate that the association between the meaning of the news articles and the crude oil price is not sufficiently strong. It is additionally concluded that the XGBoost algorithm reveal superior predictive performance in the training set. As a result, XGBoost models for each month of 2022 were developed to investigate the impact of features and determine the most important ones for the problem. |
|---|---|
| Autores principais: | Vieira, Joana da Silva |
| Assunto: | Latent Dirichlet Allocation (LDA) Ridge Random Forest XGBoost ARIMAX crude oil price forecast |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | Crude oil price volatility has an impact on the global economy and oil-dependent industries and is influenced by supply and demand, geopolitical tensions, and the global economy. Every day, a massive amount of textual information flows in the form of news articles, which humans use to forecast future trends. News articles can have a significant impact on the price of crude oil because they contain information about recent events, trends, and advancements in the industry. The purpose of this work is to investigate how news articles may affect crude oil prices, using the concept of topic modeling and its potential for handling data. Using the webscraping method, the data for the study comes from a large dataset of news articles about the crude oil industry. These news articles were published between January 1 and December 31, 2022, and come from four different sources. The data was compiled using the source Exchange Rates UK to demonstrate how the price of crude oil fluctuated during this period. After the cleaning process was completed, the dataset contained a total of 1532 news articles. The Latent Dirichlet Allocation (LDA) technique is suggested for extracting relevant keywords from news articles and then using the findings as input features to forecast the crude oil price. The forecasting methods employed in the study were the Ridge model, the Random Forest and XGBoost techniques, and the time series method ARIMAX. The outcomes of the experiment indicate that the association between the meaning of the news articles and the crude oil price is not sufficiently strong. It is additionally concluded that the XGBoost algorithm reveal superior predictive performance in the training set. As a result, XGBoost models for each month of 2022 were developed to investigate the impact of features and determine the most important ones for the problem. |
|---|
Atividades financiadas
Não foram encontrados projetos financiados.