Bayesian Extreme Value Optimization Algorithm: Application to Forecast the Rubber Futures in Futures Exchange Markets

Abstract

This paper statistically investigated the extreme event statistics, which is valid across multiple time-period trends. The Non-stationary Extreme Value Analysis (NEVA) based on Bayesian inference was conducted. The series of RSS3 future contracts traded in Japan’s TOCOM, Singapore’s SICOM, Chinese’s SHFE and Thailand’s AFET were observed during the period 1990 to 2018. By providing posterior probability intervals (5% and 95% quantiles), the NEVA offers the 30-year predictions of the four future markets, which are able to reliably describe extremes and their future contracts levels. The scenarios of the RSS3 futures exchanges inform that TOCOM provides explicitly evidence for of long-run constancy, while, SICOM, SHFE and AFET seem to be higher price volatility. Therefore, the unbiased prediction tool, especially for the RSS3 futures contracts, is very crucial for stakeholders to hedge against price uncertainty.