Picking Mutual Funds with Self-Organizing Maps

Abstract

The self-organizing map approach can be used to translate multi-dimensional mutual fund data into simple two-dimensional maps. These maps provide a significant improvement over the information that is traditionally published on mutual funds. They create a better basis for portfolio selection, for comparison of the performance of mutual funds and for the creation of benchmarks.¹ Using data published by Morningstar, we use self-organizing maps to demonstrate clear distinctions and patterns among the best rated mutual funds. As inputs we used the performance of the mutual funds, the Morningstar risk measurements, the tenure of investment managers, and the various expense ratios. SOMs simplify classification of funds; can be used for decision-support; and to provide classifications that have more meaning than simple sorted lists based on multiple criteria.