Combining Multiple Criterion Systems for Improving Portfolio Performance

A central issue for managers or investors in portfolio management of assets is to select the assets to be included and to predict the value of the portfolio, given a variety of historical and concurrent information regarding each asset in the portfolio. There exist several criteria or models to predict asset returns, which in turn are sensitive to underlying probability distributions, their unknown parameters, whether it is a bull, bear or flat period subject to further uncertainty regarding switch times between bull and bear periods. It is possible to treat various portfolio-choice criteria as multiple criterion systems in the uncertain world of asset markets from historical market data. This paper develops the initial framework for the selection of assets using information fusion to combine these multiple criterion systems. These MCS' are combined, using the recently developed Combinatorial Fusion Analysis (CFA) to enhance the portfolio performance. We demonstrate with an example using US stock market data that combination of multiple criteria (or models) systems does indeed improve the portfolio performance.