The Strong Form of the Efficient Market Hypothesis (EMH) assumes that the stock market reflects all historical performances, public information, and private information; and has been accounted for in current stock prices. The Strong Form of EMH implies that there are only lucky and unlucky investors in the stock market, and that no investor can consistently "beat" the market. What happens if that is not the case? The purpose of this project is to use sentiment from different media and news outlets to see if consumer emotion can be used to predict next day price trends. Being able to predict next day price trends with high accuracy may serve as counter argument to the Strong Form of EMH. We use a variety of machine learning algorithms and models predict the next day price trends of several publicly traded stocks from various industries. Model performance is measured by accuracy. A resilient backpropagation neural network model achieved a testing accuracy of 68% on Tesla (TSLA) stock data. While the model's performance is lower than some of the models in the literature, TLSA stock has a much higher historical volatility than indexes that have been studied such as the Dow Jones Industrial Average (DJI). The difference in accuracies may provide a foundation for further research in regard to how volatility may impact model performance.
Predicting Stock Prices and Trends Using Sentiment Analysis and Machine Learning Algorithms
Category
Mathematics