Optimal Tactical Asset Allocation – Using Recommender Systems in Portfolio Management

Abstract

Here we present a novel approach to how the Chief Investment Office (CIO) can select investment strategies to allocate to and to decide the percentage allocation to them. The method that we outline here is a continuation of our previous research on recommender systems science[13]. The aim of this project is to improve upon an objective measure of the CIO’s efficiency. The CIO office is tasked with allocating to the available investment strategies and altering the allocation based on changes in financial markets. The allocation algorithm outlined here tries to adapt to new market regimes while simultaneously responding to observations of the relative performance of investment strategies. It turns out that a solution to the strategy selection problem faced by CIOs is very similar to the science behind movie or music recommendation. The algorithm learns from historical observations about strategies. It tries to figure out a relationship between different dates and months in the past, and between strategies to which we are possibly allocating. This process of detecting a two-dimensional neighborhood is what is superior to methods adopted by CIO offices in the past.