[2I1-GS-2] Machine learning: Random forest

[2I1-GS-2-03] Exploring Determinants of Investor Sentiment in the FX Option Market

〇Kazuaki Washimi¹, Kimiaki Shinozaki¹, Yasufumi Genma¹, Yasutaka Takizuka¹ (1. Bank of Japan)

Keywords:Machine learning, Random forest, Regression analysis

With novel, granular FX option data, this paper employs a Random Forest approach to exploring the determinants of investor sentiment for USD/JPY by a different investor category. The sentiment is measured by the positioning of “long call and short put (bullish for USD/JPY)” minus “long put and short call (bearish for USD/JPY)”. The analysis shows that the uncertainty over the US trade policy is one of the most important variables for the sentiment of non-financial corporates, while the US yield curve appears to affect the sentiment of both non-financial corporates and institutional investors. The results imply that the recent increase in hedging behavior for a sharp fall of the dollar against the yen could be attributable to the US-China trade tensions and an inversion of the US yield curve which suggests a slowdown of global economy.