On a new calibrated mixture model for a density forecast of the VN30 index

Abstract

Regarding predictions of business and financial quantities, seldom has a consensus been reached among experts which, in certain cases, creates insurmountable difficulties for decision-makers to reach a final decision. In this paper, motivated by the quest for a reliable forecast of the VN 30 index, we introduce a novel method to mix and calibrate a number of stocks to better predict the index. Treating each stock as one “expert’s opinion”, we construct an integrated forecast by applying a beta calibration to a mixture model of stock historical data to derive a combined and calibrated density function for the VN 30 index. Since all the computations are carried out within the framework of bayesian statistics, our new technique is part of the bayesian semi-parametric methods.