Cuckoo search for portfolio optimization. (c2014)

Abstract

Portfolio optimization is the problem of allocating capital over different assets in order to maximize the return on the investment and/or minimize its risk, which has been a major concern for investors throughout the world. Markowitz developed the mean variance model as a part of the modern portfolio theory. When applying this theory to real-world problems, investors would force certain constraints on the solutions for a portfolio so that they meet their investment needs. Hence, the study of the portfolio optimization problem can be tackled in different ways according to the constraints to be included and the algorithms that will be applied. In this work, we design a meta-heuristic method based on Cuckoo Search (CS). Cuckoo Search is an optimization population based algorithm that imitates the reproduction strategy of cuckoos. The meta-heuristic is evaluated using different problem model parameters such as portfolio size, cardinality and lower and upper bounds. The dataset used is one provided the OR Library. The measurement of the solution quality in the results is compared with previous results and shows some improvement. For example, the CS algorithm outperformed two algorithms out of three in all datasets. In addition, it outperformed the third algorithm in 2 datasets out of 5.