Quantification of causality from sentiment to market fluctuation 

For a long time, economists did find that sentiment has, to a great extent, influence economy situation. People’s opinions (positive or negative) towards a product will influence the sales. People’s emotions (optimist or pessimist) will influence the stock market movement. In this field, behavioral economics has done a great job. Through studies of the effects of psychological, cognitive, emotional, cultural and social factors on the economic decisions of individuals and institutions. Economists divide factors for decision making into 2 parts: endogenous factors, which are factors that influence the business cycle from inside the system; and exogenous factors, which are those from the outside the system. For the former factors, big company’s monetary policy and government policy are typical examples. For later factors, sentiment is found to be an important exogenous factor. But all of these could not satisfy the need for decision making in economic activities, where people need more precise interpretation of the market, which means all the factors, sentiment, market fluctuation, the relationships in-between, should be quantified. This is just what I am doing. I’m devising models and methods to obtain quantified results in order to get valuable guidance for economic activities. These methods could also be used in other fields concerning causal inference and sentiment analysis.

Generally, the research in causal inference includes 2 tasks: “discovery” and “prediction”. Quantified methods are needed in both tasks. My overall objectives are (a) sentiment analysis which extracts psychological and exogenous factors from texts, audios and videos; (b) causal inference from sentiment to market fluctuation, where the causal inference should be quantified; (c)  a sort of recommendation system, which uses prediction task of causal inference and make recommendations.