Own sentiment analyzer applied to stock price prediction

: a case of Apple : 1996 to 2016

Abstract

Applications of sentiment analysis for the world of finance can no longer be ignored. The idea of factors beyond the economic field affecting stock prices has consequently been sustained by practitioners over the years. This topic is often referred to as behavioral finance. One of the factors outside of the scope of finance which is commonly addressed is investor sentiment. It is commonly agreed that investor sentiment is gauged by news articles. While a great amount of work is conducted on news from social media platforms like Twitter, news articles used for this business project come from well-known newspapers. This business project explores two approaches which differ from most of the existing research. Firstly, it addresses both the stock price prediction and the sentiment analyzer as regression problems, in contrast to the commonly used classification algorithms. Secondly, financial-related variables are also used as inputs for the stock price prediction, in opposition to most literature using only the sentiment classes for that purpose. A lot of time was allocated to processing the unstructured textual data in a way which would be appropriate to process the news from Apple, as this is considered a key step in sentiment analysis. In what concerns the results obtained, while Support Vector Regression proved promising for not only the sentiment analyzer but the stock price prediction, the algorithms’ performance has plenty of room for improvements. Results point to sentiment impacting in fact stock prices, at least to some extent, as according to previous work.

Date of Award	20 Oct 2021
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Viktor Pekar (Supervisor)