Stock price movement prediction using distributed representations of financial report

Abstract

In this research, we aim at predicting stock price movements based only on text data from financial reports by applying recent natural language processing techniques.　In particular, we propose neural network-based document embedding models to represent text information from the large amount of financial reports to show the possibility that natural language processing approach can be used for stock price prediction. The contribution of this research is to show how to acquire proper text representation that can be useful to predict stock price movement.