Visualization of the Market Sentiment in Stock Markets based on Neural Network

In this study, we tried to visualize the market sentiment that affects the judgment and behavior of stock market participants. Its purpose is to clarify the periodicity contained in the time series data of TOPIX and S&P500. The trend and volatility for a given period of time were calculated for price fluctuations from 2005 to 2014 and we used as training data for the neural network. We prepared three types of learning data: short, medium, and long term, according to the length of the period over which we calculate the trend and volatility. An agent with a neural network learns using one of these three types of training data, and each predicted a near-future price after the training was completed. Therefore, we used three agents in this study. By comparing the predictive results of these three agents, we attempted to visualize market sentiment.