Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 52564 (19)
- University: Sharif University of Technology
- Department: Computer Engineering
- Advisor(s): Habibi, Jafar
- Abstract:
- Stock price prediction is a serious challenge for investors and corporate stockholders to forecast the daily behavior of the stock market, which helps them to invest with more confidence by taking risks and fluctuations into consideration. Stock price prediction is a difficult task, since it is very depending on the demand of the stock, and there is no certain variable that can precisely predict the demand of one stock each day.During the past few decades, time series analysis has become one popular method for solving the stock forecasting problem. However, depending only on the stock index series makes the performance of the forecast not good enough, because many external factors which may be involved are not taken into consideration. As a way to deal with it, sentiment analysis on online textual data of the stock market can generate a lot of valuable information, which can be named as external indicators. Nowadays, social media is perfectly representing the public sentiment and opinion about current events and is a good resource for sentiment analysis.In this study, we explored data from StockTwits, a microblogging platform exclusively dedicated to the stock market. For the embedding purpose, we concatenated all tweets of a day instead of just considering each tweet individually. Then we trained our doc2vec model with different feature sizes on this data to generate a feature vector for each trading day. The building block we used in our architecture was GRU since it showed a better performance compared to LSTM. We evaluated our model on various stocks from several industries having different dataset volume. Experimental results demonstrated that our model accuracy achieves nearly 60% in the S&P 500 index prediction, whereas the individual stock prediction is near 74%, which overtook our base model. We also developed a hybrid model that used both textual and price data to leverage trade information, but the result was not promising in comparison with the previous model
- Keywords:
- Sentiment Analysis ; Stock Market ; Deep Learning ; Neural Network ; Stock Price Movement Prediction
Digital Object List
-
محتواي کتاب
- view
Bookmark