Portfolio Optimization Using CART and Genetic Algorithms

Abstract

In this article, we present an approach based on the approximation of data by segmentantion using the CART algorithm and Genetic algorithms, with a view to constructing an optimal portfolio-extract of financial assets, . This approach generates a surplus of financial gains in terms of costs, and improves performance by reducing computing loads. This is a three-stage process: the first is to represent the financial series using a piecewise linear approximation obtained by CART, where the trend change points are optimally determined by segmentation of the series into periods of economic regime change.In the second stage, once the segments have been determined, each segment is represented in the plan (yield, volatility).We then develop an algorithm that selects the top ten financial assets assets in the overall portfolio, known as the extracted portfolio, in terms of the best ratio between return and VaR. In the third step, we apply an optimization algorithm based on genetic algorithms to the extracted portfolio . The aim of this algorithm is to optimize the weights that will minimize the VaR and maximize the value of the extracted portfolio.