Forecasting Corporate Revenue by Using Deep-Learning Methodologies

In the past few years, deep learning evolved into a powerful machine learning technique, which uses multiple layers for feature representation to learn specific attitudes of the raw input data, in order to produce state of the art prediction results. Deep learning has become popular in many application domains which use rich variety of data. Large volumes of online business news provide an opportunity to explore various aspects of companies. Sentiment analysis of text establishes a new viewpoint of large scale data identifying, among other features, the tone of the author towards the subject of the text. Hence, the sentiment of news articles offers an insight into the internal state of the company, potential for revenue growth, and it can be useful for corporate decision making of the company. In this paper, we demonstrate a deep convolution LSTM neural network that uses a fusion of data including company stock price and sentiment of company-related news articles as time-series, in order to predict the revenue growth or decline of the companies belonging to the Dow Jones Industrial Average. Additionally, we present a method based on transfer learning for sentiment analysis of news articles related to finances, and compare this method with standard statistical sentiment analysis approaches.