Details

Autor(en) / Beteiligte

• Yin, Hujun
• Gao, Yang
• Chen, Songcan
• Wen, Yimin
• Cai, Guoyong
• Gu, Tianlong
• Du, Junping
• Tallón-Ballesteros, Antonio J
• Zhang, Minling

Titel

Finding Sentiment in Noise: Non-linear Relationships Between Sentiment and Financial Markets

Ist Teil von

• Intelligent Data Engineering and Automated Learning - IDEAL 2017, 2017, Vol.10585, p.580-591

Ort / Verlag

Switzerland: Springer International Publishing AG

Erscheinungsjahr

2017

Link zum Volltext

Quelle

Alma/SFX Local Collection

Beschreibungen/Notizen

• Sentiment analysis, especially in economics and finance, start-ed with a great fanfare in the late 1990’s when (negative) sentiment proxies were introduced in econometric schemes used in finance for forecasting price returns. Intuitively, the relationship between returns and the proxies suggests a polar opposition - high negative sentiment low returns and vice versa. The parametric (vector autoregression) analysis has suggested a linear relationship between two variables especially in ostensibly well-regulated and stable markets like New York Stock Exchange. We have examined this return-sentiment relationship over a period of many years in stable (NYSE), quasi-stable (Copenhagen Stock Exchange and the Chicago Board of Trade’s commodity markets) and the emergent markets (Shanghai Stock Exchange). A non-parametric method (locally weighted regression) with variations across the board has graphically shown the relationship between sentiment and returns in a non-linear fashion. Our study is largely based on newspaper report (Shanghai and Copenhagen), opinion pieces (New York and Chicago) and blogosphere (Chicago) - totalling 39.2 million tokens regressed with time series comprising 15,871 data points. Evidence from both models shows that negative sentiment actively influences stock and commodity markets respectively. The analytical conclusions determine that sentiment represents a meaningful performance in predicting momentary market-wide drops in valuation.