A carregar...
Publicação
Sentiment analysis of online media - extracting a trading signal for commodities
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 29.03 MB | Adobe PDF |
Autores
Orientador(es)
Resumo(s)
Sentiment Analysis is an extensively studied topic that is increasingly applied in financial applications. As a method to forecast asset prices, text classification is used to incorporate the sentiment of investors to enhance the accuracy of price forecasts. This thesis examines a broad range of news websites from the Common Crawl news dataset. The challenging task is applying topic models to extract sentences, targeting the development of crude oil price movementin acoherent way, by evaluationthe accuracy of the model. The results revealed that highly targeted data is necessary to make appropriate price predictions based on news sentiment. Extracting meaningful sentences is a challenging task that needs further investigation. A framework for incorporating and processing the data to prepare it to fit the text classifiers can further enhance the already impressive results of predicting the prices of financial assets based on investors' sentiment.
Descrição
Palavras-chave
Machine learning Financial markets Data science Natural language processing Sentiment analysis Price prediction Investor sentiment Text classifier News sentiment Time series forecasting Commodities Online media