Applications of intelligent systems for news analytics in finance

Auteurs

• C. aslav Boz ic

Paramètres

• Manuels
• Manuels universitaires

En savoir plus sur le livre

News articles are crucial for traders as they can trigger significant market movements, including panic selling or large-scale buying, as well as more subtle shifts. With the increase in algorithmic trading, there is a growing need to quantify qualitative information from news texts and integrate it into trading algorithms. In 2007, Tetlock's pioneering research applied empirical finance methods to media sentiment, demonstrating that sentiment measures based on the Harvard IV-4 psychological dictionary correlate with future equity prices. This approach was expanded to develop a general sentiment score, facilitating the evaluation of advanced text mining techniques. A methodology for creating benchmark datasets from historical data was proposed, using the Thomson Reuters Tick History database to establish a benchmark for financial text mining with over 570,000 data points linked to more than 100,000 news articles. This benchmark enabled a comparison of three key text classification methods: Bayesian classifiers, support vector machines, and artificial neural networks. The findings revealed that support vector machines outperformed the other methods, including multilayer feedforward neural networks. The results of sentiment analysis can assist professionals in making informed financial decisions through visualization techniques or serve as additional parameters in trading algorithms. The FINDS project specifically developed a se

Achat du livre

• Année de publication: 2013