Title:Multivariate Gaussian and Student$-t$ Process Regression for Multi-output Prediction

Abstract:Gaussian process for vector-valued function model has been shown to be a useful method for multi-output prediction. The existing method for this model is to re-formulate the matrix-variate Gaussian distribution as a multivariate normal distribution. Although it is effective and convenient in many cases, re-formulation is not always workable and difficult to extend because not all matrix-variate distributions can be transformed to related multivariate distributions, such as the case for matrix-variate Student$-t$ distribution. In this paper, we propose a multivariate Gaussian process regression (MV-GPR) as well as multivariate Student$-t$ process regression (MV-TPR) for multi-output prediction, where the model settings, derivations, and computations are all performed in matrix form directly, rather than vectorizing the matrices involved. Compared with independent Gaussian process regression and Student$-t$ process regression models, both MV-GPR and MV-TPR significantly have the outstanding performances in the simulated examples and the corresponding Buy\&Sell strategies can be also shown profitable in stock market investment. In particular, MV-TPR has a distinct investment in NetEase among three Chinese stocks from NASDAQ. From the view of industrial sector, MV-GPR has considerable performances in Industrials, Consumer Goods, Consumer Services, and Financials sectors while MV-TPR can gain maximum profit in Health Care sector.