Computer Science Department
Neural Network for Stock Trading and Text Utilization
This project entails a web-based application that will help stock traders to notice the connection of the daily news and the financial market movement. The application will scan each news feed and search an inherited database to find the most related company. The database possesses a list of companies and the corresponding neural network. Each neural network scan the news input and predicts how related it is to the company. The search engine will let the original news input go through several neural networks and find the highest output. The output will be displayed on the web, so users can see every news article with the most related stock option. The news input does not only takes a business report but also more broad and general topics.
This project has its application and goal in the financial field, especially stock trading. However, it does not use any financial knowledge or indicator such as stock price. Rather, the complexity is on the use of neural network with text information such as qualitative news articles. Each neural network will test and self-learn a sufficient amount of historical news about the company so that the network has a list of keywords subtracted from them and the weight on keywords. User interface will have an input form where user can type news headlines and an output displays.